D-Wave Quantum Aktie

Please welcome to the stage Senior Director, Investor Relations, Kevin Hunt.

Okay. Thank you, and thank you all for joining us today for D-Wave's first ever Investor Day. Thank you for all of you in the room for joining us, making all the way down to the NYSE, and thanks to all the hundreds that are listening to the live stream online. We 've got a great lineup of presentation for you today. So hopefully, you'll all be quantum experts by the end of the day.

Just a few notes on the agenda. We will be having several opportunities to ask questions during the day. Our CEO, Dr. Alan Baratz, will have -- answer some questions from people in the room following his keynote. We'll have two other places you can ask questions. For those of you in the room, please raise your hand, and we will bring a microphone to you. We ask that you identify yourself before you ask a question and that you limit yourself to one question.

Following the end of the last Q&A session at the end of the day, we will be having a reception for people out in the room -- in the back room there from 4 to 5. You'll be able to see some demos and meet with the members of the D-Wave management team. For those viewing online in the live stream, you also can ask questions via the Ask button and we'll try to get to as many of those as we can.

So turning to the ever popular disclaimers, we will be making -- members of D-Wave's management team will be making forward-looking statements over the course of the afternoon. There are always risks and uncertainties associated with any such forward-looking statements. You can see a list here in the slide behind me. And this investor deck has been posted to our company website. So you can see the list there as well. If you want even more risks, you can go to our 10-K filed with the SEC and also available on our company website.

We also will be referring to some non-GAAP financial metrics such as bookings throughout the day. And we may -- in case that other non-GAAP financial metrics come up during Q&A, you can see a full list and descriptions of those in our earnings releases, which are available on our company website.

So with that, I'll turn it over to our CEO, Dr. Alan Bars.

Please welcome to the stage, D-Wave's President and CEO, Dr. Alan Baratz.

Good afternoon. So my name is Alan Baratz. I am the CEO of D-Wave. I have no clue who that guy in the picture is. It's a real pleasure to be here with you all today. I thank every one of you who took the time to come down here to the New York Stock Exchange and be here in-person. Thanks also to those of you that are joining us via the live stream.

There's quite a bit that we would like to cover with you today. But honestly, the Q&A sessions are equally, if not more, important. We want to make sure that there's plenty of opportunity to ask questions. And so we've allocated three different times for Q&A. And hopefully, through that, we'll have the opportunity to answer most, if not all, of the questions.

So I'm going to start with the obligatory picture of Richard Feynman, although hopefully, you haven't seen this version of the picture before since our excellent graphic design has made it up. Richard Feynman, of course, coined the term quantum computer back in 1981. And he's famous for a number of other interesting and important commentary on quantum mechanics and quantum computing.

One that I think is particularly relevant to us, and by us, I mean the entire quantum computing industry here today; is the one that we have up on the screen. If you think you understand quantum mechanics, you don't understand quantum mechanics. This is very complex technology.

The problem with that is that it's way too easy to say misleading things and to spew hype into the marketplace, and it's hard for all of us to be able to separate all that noise from the signal. And so part of what I'm hoping to do here today is at least for D-Wave, help you think about how to separate the signal from the noise and how to think about D-Wave in the context of what we are telling you against, if not the science, at least sound logical thinking.

So I'm going to start at the very top. What is quantum computing? For me, quantum computing is energy-efficient computing for solving hard computational problems. That's typically not where people start when they talk about quantum computing. But I start there because quantum computers consume very little power.

And on top of that, they use quantum mechanics to compute the solutions to problems very, very quickly. And that combination truly allows them to be much more energy efficient at solving problems than classical computers.

Now there's more than one approach to building a quantum computer. And this is particularly important in the context of understanding D-Wave and how we are different from every other quantum computing in the industry. There are two main approaches to building a quantum computer. One is called annealing and the other is called gate model.

For those of you that have been following us over the years, you've heard us talk quite a bit about annealing because, frankly, we started with annealing, and we're the only company in the world today that brings annealing quantum computers to market.

As we go through the course of the day, we're going to talk not only about the power and value and road map for annealing, but we're also going to talk about gate model because we are now also a significant player in gate model, quantum computing.

But before I dive into it, let me just take a minute to try to explain the difference between gate model quantum computing and annealing quantum computing. And I'm going to start by talking about classical computers.

So typically, you'll hear people say that classical computers use to store information, and a bit can be either a 0 or a 1. And this means that at any point in time, the information that the computer, the classical computer is seeing is one possible solution to a problem. It's how 0s and 1s have been assigned to the various bits. That shows you one possible solution. It's like dropping a tennis ball into a landscape, and it hits at one point.

Now suppose you want to find the overall lowest point in this landscape, which is the valley to the left of this image, well, how do classical computers work? They basically perform computations and logical operations that attempt to explore the landscape one point after the next. They're moving that tennis ball around point by point, looking for the lowest possible point.

Now if your landscape is simple like a bowl, it's easy. There are very efficient algorithms for finding the lowest point. You can get there quickly. But if it's a complicated landscape, this is actually an exponentially hard problem. What it means is there's not enough structure in the problem to be able to efficiently find that lowest point. The amount of compute time is exponential in the size of the problem.

And this is exactly the kind of problem that's good for quantum computers, the very hard computational problems. So how does a gate model quantum computer work?

Well, gate model systems use qubits, and a qubit can be 0 and 1 at the same time, which means that at any point in time, you're able to see multiple possible solutions to the problem, not just one possible solution, and you're able to perform operations on those multiple possible solutions simultaneously, computations, logical operations.

So the problem becomes one of not kind of figuring out how to follow the ball to the lowest point, but you've got all the balls. You can see all the points. Now the algorithms are all about figuring out which one of those balls is actually at the lowest point. It's a different way of thinking about computation and how to program the computer.

But there's one other important point about gate model systems, and that is quantum mechanics was not really designed to support digital computation. So as we build these gate model systems, we're trying to implement digital computation on quantum mechanical systems.

We're pushing against nature, and that introduces errors. That's why gate model systems are so error prone and why without the ability to correct those errors, we're never going to be able to solve useful problems on gate model systems. So error correction is critically important for gate model systems.

Annealing quantum computers work in a very different way. Annealing quantum computers are really all about finding the lowest point in a large landscape. They're about using quantum mechanics to find what we call a low-energy state. It's working with quantum mechanics.

You can think of it like dropping water into the landscape and looking at where it pools. It pools at the lowest possible point because the water is able to tunnel through the hills and valleys and get to the lowest possible point.

Annealing quantum computers use quantum mechanics to do exactly that. They use superposition, they use entangle and they use tunneling. They work with nature to find the low point or solve the problem, right? And as a result, they're nowhere near as sensitive to noise and errors. And that's why our annealing quantum computers are solving hard customer problems today without the need for error correction.

But both are important. There are problems that require annealing, specifically, business optimization problems are well suited to annealing. Gate model systems are not very good at solving them. And there are problems that require gate model, quantum chemistry, for example, materials discovery. These are well suited to gate model systems, not very good at solving optimization problems, right? You need both annealing and gate in order to be able to address the full market for quantum.

Okay. So now how do we think about finding the signal within all the noise that's going on out there? Well, we think that to be successful in the quantum computing industry, you need three things.

One, you need to be able to address the full market for quantum, all the possible use cases. Two, you need to do that with viable commercial products, not research tools, but viable commercial products. And three, you need to be able to deliver measurable customer value in using those systems, address the full market with commercially viable products, delivering measurable customer value.

And we think that D-Wave is right now pretty much the only company in the world, the only quantum computing company that can address and deliver on all three of those, and I'm going to explain why.

Let's start with the first, address the full market. Okay. So there are many different TAM analyses that have been done by different consulting firms, TAM, total addressable market analysis. Boston Consulting Group has done it. McKinsey has done it, a variety of other consulting firms have done it.

But the one that most of the industry has latched on to for 5 years now is the Boston Consulting Group study. They put it out initially about 5 years ago. They refreshed it about a year ago.

BCG puts the total addressable market for quantum at just under $1 trillion in roughly the 20-year time frame. And they divide it into 4 different technological segments or use case areas: cryptography, optimization, machine learning or AI and simulation. And they assign a value to each of those adding up to the $850 billion total value.

Well, if you think back to what I said a few minutes ago with respect to annealing and gate model, gate model cannot address all of these, annealing cannot address all of these. Gate model, for example, is required for quantum mechanical simulation, drug discovery, materials design. It's also quite good at breaking codes, right? Annealing quantum computers are required for optimization problems, workforce scheduling, manufacturing plant floor optimization.

You need both to address the full set of use cases for quantum computing. And D-Wave is the only company in the world bringing both annealing and gate model to market. We're the only company in the world that can address the full TAM for quantum, the full set of use cases for quantum.

In addition, sometimes the gate model companies will have optimization envy. They vacillate between "Oh, we can do optimization as well. Look at -- we solved this optimization problem as well as it can be solved on a laptop. See, we can do optimization as well." Okay, as well as it can be solved on a laptop, what about solving it better than it can be solved classically? There's significant, both theoretical and experimental evidence that gate model systems just can't beat classical on optimization, but annealing systems can.

Then sometimes we'll hear, "Oh, annealing is just a niche. It only addresses a niche market." Well, is optimization a niche market? First of all, it's a full 25% of the TAM that we just discussed. But think about it, the numbers, $100 billion to $220 billion, bigger than the global semiconductor market, bigger than the global cybersecurity market. Are those niche markets? I don't think so.

The point being that there's only one quantum computing company that can address the full TAM for quantum. And each of our products, our annealing quantum computers and our gate model quantum computers, are addressing a very significant portion of the TAM.

Okay. Let's move on to commercially viable products. I want to start by defining commercially viable. You will hear every quantum computing company say, "We're commercial today. We've sold a quantum computer to this company or this company is using our quantum computer to work on this collection of problems," right?

Research experimentation with a quantum computer is not commercial use of the quantum computer. Commercial use of the quantum computer is using it as a part of your business operations on a daily basis, an integral part of your business operations. That's what it means to be commercial.

So again, lots of noise, lots of hype, "Oh, we're commercial." Next time, your favorite quantum computing company other than D-Wave because I know D-Wave is all of your favorite, but next time your other favorite quantum computing company says, "Oh, we're commercial," or say, "Who's the customer? What's the application? How often are they running it? Can I go talk to them?" ask us, we'd be more than happy for you to talk to our customers.

Every time we have an event, we actually have customers participating in the event with us. So you've got to cut through the noise. Commercial means in use on a daily basis as part of business operations. Now the customers that we bring on stage, some of them are commercial, some of them are still in the R&D phase. And we're clear to talk about that as we engage in those discussions.

Okay. So let's start with our annealing products. Our flagship annealing quantum computer is our Advantage2 system. It is a 4,500 qubit annealing quantum computer. We've also built a complete ecosystem of products around that. We have a quantum cloud service called Leap. This is designed to support business applications in production, building in the reliability, availability, security, privacy needed to be able to support commercial applications.

We also have a complete suite of software development tools, our Ocean SDK for building applications, leveraging our annealing systems. We have a suite of hybrid solvers that bring together classical and quantum to solve problems on the quantum computer that are larger than can natively fit onto the quantum computer.

And we have a professional services capability for helping our customers understand which applications can most benefit from quantum and how to build out those applications. This is a very comprehensive and very mature platform of products in the annealing arena.

We've got one of the largest aerospace companies working with our products, one of the largest airlines working with our products, one of the largest chemical companies working with our products, one of the largest mobile carriers working with our products. We're working with government contractors on national security and defense.

This technology space is quite mature today. Still more work that needs to be done, but nonetheless, quite mature today. And we've talked about some of the metrics associated with improvements that these customers have seen leveraging our quantum computers. We'll talk a little bit more about that later.

We've also pointed out that our annealing quantum computer, our Advantage quantum computer is the only quantum computer in the world that has been able to solve a useful real-world problem that cannot be solved classically, true quantum supremacy.

We published this in science a little over a year ago. This is computing properties of materials that were computed on our system in 20 minutes, that's the image on the right, and it would take nearly 1 million years, billions of minutes to solve it on the fastest supercomputers in the world.

Now I'm going to just diverge for a minute and talk about a little bit more hype, right? There are many people that would love to say we have not achieved supremacy on this system, and they try. And you may have seen a headline a week or 2 ago that a group of researchers just published a paper in science where they basically have shown that classical can solve the problem that we solve.

Okay. Go read the paper, right? The headline and the paper are very different things. They have not solved all the lattices that we solved, all the sizes of the lattices that we solved, all the evolution times that we solved and all the properties that we computed, all of which is required to achieve supremacy. They've done a small subset of what we have done.

This supremacy result has been out there for almost 2 years now. It was published in science a year ago. It was put on the technical archive a year before that, and it still stands. This is the one true demonstration of quantum supremacy.

As far as our annealing roadmap is concerned, we've talked about the fact that our Advantage3 system will have 100,000 qubits. We haven't really put dates around that previously, but now we are providing dates for the 100,000 qubit Advantage3 system and a smaller 20,000 qubit version.

There are two important technologies that are enabling our move from 4,500 qubits to -- or 4,400 qubits, sorry for rounding up, to 100,000 qubits. They are multichip, basically moving from more qubits on a chip to interconnecting processor chips, multichip and scalable control.

On the multichip front, basically, what we're talking about, and it's shown here, tiling in 2 dimensions, our 4,500qubit processor chips. So 4 to 5 chips to get to 20,000 qubits, 25 or so chips to get to 100,000 qubits. We need to be able to interconnect these things electrically through bonds that are superconducting and preserve quantum mechanical properties.

This is work that we've been doing with the NASA Jet Propulsion lab. We've developed the materials and the process for this, and we've actually been able to demonstrate the ability to bond the chips and preserve these properties. So that was one of the two technical challenges that we needed to overcome to drive to the Advantage 300,000 qubit system.

The second, scalable control. And this is a really important point because it's going to come back when we talk about the gate model systems as well. So typically, if you go ask anybody in the industry, how many control lines do you need to control a qubit, you'll hear 3 to 5 per qubit. So that would mean that our 4,500 qubit Advantage2 system would require 12,000 to 20,000 I/O lines.

We control that 4,500 qubit processor with 200 to 300 I/O lines because we've been able to build data pipelining, multiplexing, addressing onto the processor chip, which means we can much more efficiently control the qubits. And that's critically important.

Now when we go from 4,500 to 100,000 qubits, though, we can't have a 25-fold increase in the number of I/O lines. So we've been working on modifying our control architecture to make it even more scalable than it has been up until now. We've completed the design. We've completed the masks for the first chips to basically test out the design, and they're going through fabrication now. And that's the second important component of being able to scale to the 100,000 qubit processor.

Both of these technologies are well down the development path. But obviously, we have some time and work to test and really kind of ring out any issues associated with either of these technologies. But we're quite enthusiastic about the path to 100,000 qubit.

And why is this important? So if we think about our current 4,500 qubit system, when combined with our hybrid solvers, we can support problems with up to 2 million variables. And that's fine for solving a lot of problems. But if, for example, we wanted to solve full-up last-mile routing for FedEx or UPS, that we require 50 million variables. So we still have to continue scaling on the annealing side to solve larger and more complex problems.

All right. Now I want to spend a little bit of time talking about the gate systems. We announced quite a few years ago that we're going to start working on a gate model quantum computer as well. A lot of the work that we had been doing was looking at how you could take that control architecture that I talked about for annealing systems and move it into the gate model environment.

We have not spent as much time on high-quality qubits and the ability to error correct those qubits. But what you need in order to have a commercially viable gate model quantum computer is you need error correction and efficient error correction, you need scale and you need for the systems to be fast, right? Nobody wants a slow quantum computer.

And so we've been focused on how to advance the state-of-the-art and our technology development in each of those areas, efficient error correction, scalable systems, fast computation. So we are bringing together to solve this collection of problems two critically important technologies.

Relative to the ability to efficiently error correct and have fast computation, we closed the acquisition of Quantum Circuits in January. This was a company that was spun out of Yale University and founded to basically commercialize some very important new qubit technology advances that have been developed at Yale. And in a few minutes, you're going to hear from Rob Schoelkopf, who was leading all of that work at Yale and was the founder of Quantum Circuits.

The qubit technology that was developed at Yale, spun out into Quantum Circuits, acquired by D-Wave; is very revolutionary. First of all, it is superconducting qubit technology, which means it is fast. Superconducting qubits run 1,000 to 10,000x faster than trapped ion or neutral atom qubits. These are superconducting qubits. They're very fast.

But typically, the trapped ion or neutral atom folks won't argue that they're slower, but they'll say, "But our qubits are more natural qubits. They're more efficient qubits. They're much easier to error correct. Superconducting may never be able to error correct."

Well, here's the really important thing about the new qubit technology called dual-rail qubits. They have the same efficiency as trapped ions or neutral atoms. So think about it. These are revolutionary qubits that have the speed of superconducting and the efficiency or reliability when it comes to error correction of trapped ions or neutral atoms, the best of both worlds.

Then we are going to marry that with the control technology that we had developed for our annealing quantum computers to allow us to much more efficiently scale these gate model systems, essentially attacking all three of the elements required to truly have a commercially viable gate model system, efficient error correction, scalability and speed.

Think about it this way. Trapped ions or neutral atoms are like a bicycle. Simple and efficient, easy to error correct, but slow. Superconducting, the things that IBM or Google are using; very, very fast, but very complex, not all that reliable, kind of like an old piston airplane. D-Wave's dual-rail qubits are the best of both worlds. They are fast, but they are also simpler and far more reliable, like turbine jet engines, the best of both worlds.

Okay. Let me just provide some metrics now to help back this up. We think that there are two metrics that it's really important for everybody to start focusing on when it comes to getting to truly commercially viable gate model quantum computers. One is speed, and the other we're calling lambda. This is the error correction efficiency. And I'm going to spend a minute talking through this.

First of all, speed. The chart on the left shows you the time required to perform an error correction cycle on superconducting, trapped ion and neutral atom. Look at that. 5 microseconds for superconducting, 30,000 microseconds for irons, 200,000 for neutral atoms. They are very, very slow technologies.

Okay. So superconducting's fast. And I think everybody accepts that, right? The debate really has more to do with, "Yes, but can you actually error correct these superconducting qubits?" So that's where we think lambda becomes very important. Lambda is the rate at which you can reduce errors as the error correction code increases.

So if we look at the chart on the right, okay, at the top, I have surface code distances. Surface code is the typical code used to error correct gate model quantum computers. And there are 4 surface code distances shown there, distance code 3, 5, 7, 9. And typically, the number of physical qubits you need is the square of the distance of the surface code, okay, surface because it's got a length and a width, okay?

Okay. So at distance 3 -- okay, 9 we're saying 17 physical per logical qubit. Distance code, distance 5, 25, 49, there's a factor of 2, which I'm kind of glossing over; 7, 49 times 2, 97, 9, 81 times 2, 181, okay? So distance 3, distance 5, distance 7, distance 9 are increasingly larger surface codes for doing error correction. And the larger the code, the better the error correction.

At the bottom is the number of physical qubits that you need per error-corrected qubit -- logical error corrected qubit for each of those distance surface codes. Now look at the 2 lines. Currently, in the industry, we're looking at a lambda of 2. What this means is that as you increase the surface code distance, you half the errors. So you get 1/2 the errors for each increment in the surface code. That's the gray line, okay?

What we are looking at with our dual rail qubits is a lambda of 10. With each increment in the surface code, we get a factor of 10 reduction in errors. This becomes really important if you think about what you need to actually be able to use a gate model quantum computer. You need to be able to perform computations. And you need the errors to be low enough that you can perform all those computations before an error emerges.

We think that we're going to need about 1 million computations to start doing useful work. This means we need an error rate of 10 to the minus 6, 1 error in 1 million in order to be commercially viable. So now look at those 2 lines. At a lambda of 10, you can get to 10 to the minus 6 with a distance 9 service code, roughly 100 to 200 physical to logical qubits. That's the path we're on with our gate model program. That's very viable, right?

The gray line where the industry is right now, what would it take to get to 10 to the minus 6? We're talking 1,000 or more physical qubits per logical qubit by the time that gray line gets out to 10 to the minus 6. okay? That's the power in the dual-rail super qubits. They are superconducting, fast with a very rapid decline in error rate as we increase the error correction.

So for our product roadmap for gate model, we've divided it into 2 errors, what we call the post-NISQ era, where we start seeing some error detection capability that we can use in the computation; and then the fault-tolerant era, which is where we're actually at the point where we can really perform up to 1 million computations and do useful work.

So this is what you can expect from D-Wave over the course of the next roughly 6 years, with the goal being that in 2032, we would have 100 logical qubits built on 10,000 to 20,000 physical qubits with an ability to support up to 1 million gates or computations, actually the beginning of commercial viability.

So this is what you need to be asking others about. "Well, what's the speed of your computation? How rapidly are you going to be able to reduce errors as you increase the error correcting code? And when will you be able to support at least 1 million computations?" Those are the key elements, okay?

The only other thing I'm going to say about the products is we were really, really pleased to be a part of the U.S. government announcement a week or so ago, basically talking about the fact that the U.S. government is providing D-Wave along with a number of other companies, funding to basically support R&D work around the development of quantum computers.

This was really important for D-Wave for two reasons. First of all, this was the first time the U.S. government had actually endorsed D-Wave. And secondly, they not only endorsed D-Wave, but they endorsed both annealing and gate. So now the U.S. government has finally started to realize annealing is important alongside gate. They are both important. And this was an endorsement of both of those approaches to quantum computing.

Okay. Now what I want to do is transition and start talking about customers and customer value. And I want to start by, in a minute, asking Lucus Haugen to join me up here on stage. Lucus is from AT&T. To be fair, AT&T is not in production today, but they are working with our quantum computers on a variety of applications.

Lucus is an Innovative Director of Data Science at AT&T's Chief Data Office, where he leads a world-class team of data scientists, ML engineers, developers and solution architects to deliver enterprise-scale impact. He's recognized as a leader in digital transformation, leveraging agentic AI, advanced machine learning, automation and big data engineering to drive operational excellence and create business value approaching $1 billion annually for AT&T.

Lucus and his team began exploring D-Wave's impact on their operation at AT&T less than a year ago, and he's here to talk a little bit about that experience. Lucus?

Okay, Lucus. So maybe -- yes, go ahead.

Hello, everyone.

So maybe you can start by telling us how AT&T is working with D-Wave quantum computing technology.

Yes, sure. So again, we're kind of getting started. We've been working on it for a little bit. We're kind of focusing on two main problems. Of course, we're working on tackling traditional optimization problems that we're doing with classical compute today. And then we're also trying to adapt the solvers to do kind of tackle our high-intensity compute problems to see if we can get a little bit of a quantum HPC advantage out of that.

Can you maybe talk a little bit about some of the use cases you're exploring or planning to explore?

Yes. So basically, we're looking at pretty much use cases across the full spectrum or all facets of our network operations front. Some examples are going to be our technician routing optimization. That's a big one, of course.

We're also looking at -- another big one is identifying network outages and then managing those. How can we make them more efficient and more quickly? We're moving into network build optimization and even some of the network traffic optimization is kind of the use cases we're exploring.

And how has it been to work with D-Wave?

I will say that I think one of the, at least in my opinion, understated benefits of the D-Wave solvers is how easy they are to use. You take my team, a group of data scientists and engineers and myself, and we're onboarding right away with it. We've got a little bit of -- basically, anyone with a data science background who used to cranking through algorithms can pick them right up. You don't have to have a physics background.

The other thing, this is very profound as well. Because you mentioned the Ocean SDK, we can do much of the, I would say, heavy lifting, but much of the upfront work is in our own environment. And when we take the quantum use, we're basically just sending off a little array, I say little, right? It's simple, but very big array of bits or points in 2-dimensional space. That's derived from our problem set. It has nothing to do with AT&T data. It protects all of our private sensitive IP. Nothing goes to D-Wave.

Not to imply that it's not secure, it's just -- I'll tell you, it takes a tremendous burden off of my team to not have to go to governance -- data governance and security and it's safe, it's safe. We're not sending anything. We're literally sending nothing because we don't have to. I'll also point out that every time we've needed it, your solvers have been available. That's a massive advantage in kind of our space where we're cranking through quite a bit. So overall, it's very easy for my team and me to use.

Yes. I actually just want to reinforce the security comment for a minute. First of all, we have SOC 2 Type II compliance for our platform all the way through to the quantum computers. We focus heavily on security and privacy.

But the point you made is really important. It's not that you don't send things, you actually send a lot of data to our systems, but what you're sending is a matrix and a vector to set of numbers that cannot be reversed engineered into any user data or even the problem that you're trying to solve. So it's a very opaque data transfer that provides protection.

And we can't stress that enough because I don't have to worry about masking data encrypting data or anything of that nature. All it is, is a drive set of numbers from our problem set, which is meaningless otherwise without that context.

Yes. So as you look to the future, where do you see the opportunities for AT&T with quantum computing?

I think I touched on it a little bit. I think for us, it's two profound opportunities with D-Wave. Again, we're kind of looking at with the quantum tunneling, [ abatic's ]. It's one of my favorite words, by the way. The architecture itself is to optimize what we're doing today. We can take what -- in seconds we can accomplish what's taking us hours and days to do with classical optimization problems.

The second thing is we've got a tremendous -- we probably crank through hundreds of billions of signals in our real-time data use cases. And what we're trying to do is leverage the solvers to offset a lot of that. And so basically kind of twisted into kind of a high-performance computing thing.

So with optimization, I think we've got an application. We've already cranked from an hour down to less than 15 seconds. And then we're looking at exploring the HPC aspect. We can see with kind of preliminary results, the potential for magnitude of order of improvement in what we're seeing today just in our speed and efficiency.

So maybe just one last question. What advice would you give to other companies who may be curious about or interested in quantum?

That's a good way. I think upfront, and you kind of touched on this as well, is don't be afraid of the, I guess, the aspects of the quantum physics part of it. You don't really need the background to work with it.

Also, having said that, be careful that quasi-echo chamber of experts and pundits and academics saying it's not real, it's not available. It's not ready yet. It is, give it a try. It's pretty profound.

For us, I think it's just a matter of -- again, we don't have a background in physics. I can't explain annealing perfectly, but we can use it, and it's a huge lift for our operation. Most importantly is knowing how to formulate your problems in your operation, your domain. That's the trick, honestly.

Yes, exactly. Lucus, thank you. Appreciate it. And in just a minute, when Lorenzo comes up, he'll talk a little bit about how we work with customers to help kind of evaluate problems, understand which can benefit from quantum and how to build out proof of technology or proof of concept to really validate them. But thank you for being here, Lucus. Appreciate it.

So I'm pretty regularly now asked about the relationship between quantum and AI. I guess, because everybody has kind of got AI on the brain these days. So let me just spend a minute talking about my view on this.

I think that AI and quantum are very synergistic. And they work well with one another in two ways. One, they're each good at different sorts of things. and you can use them together to solve a problem, letting each work on the part of the problem that it's best at addressing.

For example, you might use AI to predict product demand for the future and then use quantum to optimize the supply chain to meet that demand. So here, you have the two technologies working side by side to solve a problem, each addressing the portion of the problem that it's best at solving.

The second, quantum can help make AI model training and inference better, faster and more power efficient. We have been working for a little over a year now with a Japanese pharmaceutical company called Shionogi. They essentially trained a model to generate molecular structures for new human drugs. And they trained the model classically, and they were getting good results.

They retrained the model using our quantum computer. And what they found was that 10x, 10x more of the structures generated were well suited to human drugs. In other words, the model was more accurate. It was also trained faster and consumed less power.

So here, we have a situation where we're actually using quantum to make AI even more efficient and more powerful. And I think over time, AI and quantum are going to come together in both of those ways, the latter potentially very transformative.

And then finally, I wanted to spend a minute on blockchain. There's some very interesting work going on right now in the quantum classical blockchain arena. We've been doing some work with a small company called PostQuant Labs. They've developed a testnet for a new approach to proof of work. The proof-of-work computation can be done either classically on CPUs and GPUs or it can be done using a quantum computer.

They currently have over 18,000 people mining in the testnet and 1,600 compute nodes on the net. Those compute nodes are mostly CPUs and GPUs and then our advantage to quantum computer.

Now our quantum computer is not online all the time. We put it online for about 5 minutes a day. When it goes online, it wins almost all the blocks. In other words, it's much faster at doing the mining. It's also more power efficient. And so what we're trying to do here is demonstrate on the one hand, using a problem that can be solved by either quantum or classical, the value that the quantum computer brings over classical, another example perhaps of quantum advantage, right?

But in addition to that, showing that if you use the quantum computer, we can actually do the mining more efficiently and with lower power consumption. And by the way, it's quantum safe. In other words, a quantum computer can't break the system.

So this is an interesting new application area for us. We're quite excited about the early results we're seeing. We have started a more substantial benchmarking phase to really understand the difference between the classical and the quantum on this test net. And at some point in the not-too-distant future, we will publish those results.

Okay. So before we kind of do the first Q&A session, let me just wrap up by going back to Feynman and just reiterating. Quantum mechanics is hard. And as a result, it's easy to spew nonsense and convince people of things that just aren't real. And you always have to challenge. As I said, customers, who's the customer? How are they using it? Can I go talk to them? Scientific results, we've been able to solve this optimization problem better than classical.

Well, okay, where is the paper? Let me go read the paper and see, are they really doing that? Or is it kind of like, well, we can solve it as well as laptop can solve it, not better than classical. Always challenge.

So for us, we are focused on addressing the full market for quantum, every use case by bringing both annealing and gate model systems to market. We've proven our ability to deliver commercially viable products through our Advantage to annealing system and the technological platform that we have built around that, and we're bringing that expertise into the gate model world as we speak.

And we actually have customers that are using us today as a part of their business operations. In Q1, we announced a Fortune 100 company had signed a 2-year $10 million Quantum Compute-as-a-Service agreement for us. I can tell you, they have moved their first application in production. So it is now running on a daily basis in production, and we are working with them on several other applications. This was our first, if you like, enterprise site license.

With that, I would be happy to take some questions. Wow. Okay. Let's go to the very back of the room.

Alan, it's Krish Sankar from TD Cowen. A quick question on the dual-rail technology. At what point would the technology incorporate D-Wave's design compared to the current QCI technology it's using? And in the future, would you move from transmons to fluxonium qubits or any decision on that?

So QCI had a 3-year roadmap that was 8 qubits, 7 qubits, 49 qubits, 181 qubits; which is a road map that we strongly believed in, and that is the plan for the next roughly 3 years.

What we are doing now is working across the teams on what happens after 2028, how do we get to the 1,000 physical qubit, 10 logical qubit processor in 2030 and the 100 logical qubit able to process 1 million operations in 2032. That's where we'll start to see the technologies between what was D-Wave and what was Quantum circuits really starting to come together.

As far as transmon versus other, I'm going to wait until Trevor and Rob talk about the roadmap, and maybe they'll call on you to ask that question to [ Ken ].

Matt Cimaglia, Quantum Coast Capital.

I know you, Matt.

Yes, you do. So a question for you. How has the -- what was the impetus essentially for the move to Florida? And also, how has that been going along so far and some of the energy behind it? And I think lastly, the FAU Advantage2 system, what are some of your hopes and dreams out of that?

It's almost like Matt is a plant. So I will tell you, the first -- Matt connected with me like a year plus ago and said, "You should really move to Florida. Let me show you around." And so we got in a car together, drove around Palm Beach. He showed me all the different areas. So that's -- in sense is where it all started.

Look, we are well down the path of getting the headquarters facility up and running. Our goal is to have that up end of September, early October. And then the R&D center will take a little bit longer, hopefully, before the end of Q1 of next year. So we are well down the path of moving the corporate offices as well as getting the R&D center set up. And we're quite excited to be there.

And the relationship with Florida Atlantic is very important. I mean, interestingly, it's not just Florida Atlantic. One of the things I really liked about the environment in Florida is the universities there all work together.

So we may have a relationship with Florida Atlantic, we may be putting a quantum computer at their site. But they're actually a conduit to all the universities in the area and frankly, a variety of other really important high-tech companies that we can work with. So we can't get there fast enough, and we're really excited about it.

John McPeake at Rosenblatt Securities. Good to see you. So could you, not a technical question, but talk about the importance of on-qubit or on-chip qubit control, what kind of ratios you can get? Because I think you mentioned the 3 to 4 that the competitors use, and the scalability issue is really big as you get out with more physicals.

Yes. Okay. So it's -- sometimes we call it cryogenic control. In the annealing processors, what we were able to do was put control logic, not so much the actual control, but more how the control signals get down to the qubits so that we didn't have to have massive numbers of I/O lines coming in and lots of electronics.

We're able to put active components on the same chip as the qubits to be able to facilitate that. And so that dramatically reduced the control infrastructure, the control costs, the I/O lines that were going down into the chip. And with annealing, we are actually able to put that all in the same chip as the qubits.

With gate model, the qubits are far more sensitive. They're much more fragile than the annealing qubits are. And as a result, we don't think we can put that on the same chip as the qubits. But that bonding technology I talked about for the multichip annealing processor, that we're going to reuse to be able to have a control chip that is bonded to the qubit chip, right?

So that we still have the cryogenic control. We still have the control in the refrigerator. We're still able to dramatically reduce the control infrastructure and the number of I/O lines going down to the chip, but it's not all on the same chip. It's control chip processing chip bonded together.

So I'll probably say something that Trevor -- they'll make Trevor and Rob go [ square ] me in their chairs. But our goal is the same thing that we've been able to do for annealing. I mean, we say orders of magnitude, fewer, right? I mean, for me, if we did square root of N for N cubic, square root of N control lines, I think that would be great. But that is orders of magnitude fewer.

Antoine Legault from Wedbush Securities. You have some important customers in Pattison Food Group, Ford Otosan in Turkey. What would you need to achieve to replicate what you're doing with, say, Ford Otosan with Ford and GM and the big 3 automakers in Detroit? Like is it once you reach the 20,000 qubit annealing system or 100,000 qubit? Just give us a sense of what you need to do to unlock larger customers in the future.

Yes. So moving from one automotive manufacturer to another isn't really so much about a larger system. It's about convincing them, a, that the system can solve their problems. Now as customers that are seeing good results with our systems become referenceable, that helps that piece of it.

The other piece is, honestly, integration within their IT environment, right? So the biggest challenge to getting something into production is getting the data that's needed to formulate the problem to be sent to the quantum computer. And that ends up being, frankly, most of the heavy lifting.

I mean the customers that have moved into production, we got through the proof of concept and the benchmarking in maybe 3, 4, 5, 6 months, and then it could be another year after that to get the IT environment and the infrastructure set up to be able to accept the problem. As we go to larger companies, that becomes harder. And that's really the bottleneck. It's not so much about the technology.

Craig Ellis, B. Riley Securities. Alan, thanks so much for doing the event. I know it takes a lot of work. The question goes back to the point you made about U.S. government recognition and the award. You've talked for years about your view that the government didn't appreciate what the company was doing, and now it does, clearly.

So the question is this, what are the specific opportunities that open up for D-Wave? Then, are there commercial businesses that might not have been as open to you without that endorsement that now are? And what does it take to execute there?

Okay. So first of all, as important as the participation in the CHIPS funding was with respect to the U.S. government endorsement, -- the thing that more caused our phone to start ringing off the hook or whatever they call it, that brings back an old image of a telephone; was when Anduril participated in our customer conference back in January. That really changed the view within the U.S. government.

And we created a new organization to go after U.S. government opportunities headed by Jack Sears, and he's got a very significant pipeline now of opportunities across many different parts of the government, mostly Department of War and Defense, but some broader than that as well.

So I think maybe even the CHIPS funding was in part as a result of that. I don't know that to be a fact, but that was very transformative for us, the fact that we were able to be successful with that use case and that they were very willing to talk about it.

Maybe one last question. Anybody on this side of the room?

Matt Ross, an individual investor. Do you see D-Wave or any other quantum companies rather following the annealing path once you show what you're doing and it's successful, do you think someone else is like, oh, we should have done that?

Yes. I love your shirt. So -- and we didn't make that shirt. So there are some companies that are starting to experiment with it, very small, like 4-qubit. There are some neutral atom companies like QuEra that are trying to explore running an annealing protocol within their neutral atom processors. But here's what I'll say. First of all, trying to do annealing within a neutral atom or trapped ion or any gate model processor is a loser. It's kind of like if you want to do gaming on your system. Do you want to run graphics software? Or do you want a GPU, right? That's the difference between trying to actually get annealing to work on a gate model system versus having an annealing processor. So I don't think those approaches will ever be truly competitive to what we're doing with our annealing processors.

As far as some of the attempts to build annealing systems, I think we could start to see more of that, but you need to keep 2 things in mind. First of all, we've got a 20-year head start, and we're not standing still. And we had to solve a lot of really hard problems to get to this point. Second, we have a huge patent moat. I mean, we've been patenting a significant number of patents every one of those 20 years, right? Many of them still active. So it's going to be really hard for anybody to compete with us in the annealing space. Okay. Thank you all. I think next is Trevor.

Please welcome to the stage Chief Development Officer, Trevor Lanting.

Good afternoon, everybody. It's great to see so many people here in person, and I understand there's several hundred people that have dialed in via online via Zoom. My name is Trevor Lanting. I am the Chief Development Officer at D-Wave. I lead our research and product development teams. I want to thank all of you for being here in person. I want to thank the people who are calling in online.

Over the course of this session, me -- I, along with my colleague, Dr. Rob Schoelkopf, are going to do a deeper dive on our product road map, both on the annealing side first, and then Rob will come up to the stage and talk about our gate model development path. Alan gave a lot of details on why we think we're going to be the first to deliver fault tolerant gate model systems. We're going to talk about really the exciting applications that we're unlocking with our annealing systems and where we're most excited in terms of applications for the gate model space and really hope to give you a sense of where we're going with the technology over the next several years.

Our mission at D-Wave from the beginning and continues to be developing quantum computing technology to solve our customers' hard problems and making that technology available to our customers as soon as possible. We have a dual platform approach to building technology. Alan spoke a little bit about annealing and gate model technologies and how they're complementary. Annealing addresses a wide variety of use cases. Fundamentally, quantum optimization is one of the killer apps of annealing technology, but there are some use cases or applications that annealing is not going to be able to address effectively.

Conversely, gate model systems will address a wide range of use cases. We're very excited about quantum chemistry as one of these early application areas. But gate model systems really -- and I'll talk a little bit about the growing experimental, empirical and theoretical evidence that gate model systems will not be able to attack optimization problems.

We chose to start our development platform, our development strategy with annealing systems now 15-plus years ago for several reasons. First, engineering the qubit controls for annealing systems is quite a bit easier than for gate systems. So this has allowed us to scale up our annealing systems quite a bit faster, and we're now at the scale of over 4,400 qubits in our Advantage2 annealing systems.

The second reason is that annealing systems are quite a bit more resilient to errors. Alan spoke a little bit about the models, the fact that we are driving down to minimum energy or low energy landscapes. We're actually exploiting the natural tendency of systems to want to be in low energy or ground state configurations. So we're working with nature and with quantum mechanics rather than against it.

And finally, and probably most importantly, there's a natural set of use cases broadly in that we call the class of problem called combinatorial optimization that map very naturally on to this process of trying to find a minimum energy in the landscape. Optimization problems occur everywhere. Alan spoke about some of the use cases, some of the customers that are getting value solving their problems today. Optimization, at its core, you're posing an objective function and you're trying to minimize -- come up with a variable combination that minimizes that objective function.

For example, you're an airline, you're trying to map flights to routes, crews to flights. You want to do this in a way that, say, minimizes your fuel cost or minimizes the downtime or the maintenance activities. This is a hard computational problem, and it's combinatorial. And when I say combinatorial, I mean, when you start adding variables, if you're an airline and you start adding routes, new cities to your overall network, the search space that you have to search over blows up quickly. It's not just a marginal increase in complexity, but it blows up combinatorily. And this is really where classical computing techniques and technologies are starting to struggle.

And then there's a class of problems between annealing and gate model system where we think that depending on the nature of the use case or the specific use case, either annealing systems or gate model systems, we'll be able to attack. And broadly, this is in the class of cryptography or linear algebra problems, machine learning problems. And I'll speak about some exciting research and exciting work that we're doing in the machine learning space with our annealing systems in a few minutes.

Our development strategy for our products really falls along this dual platform strategy. We are driving intensive investment in our annealing portfolio. Our current flagship annealing systems are Advantage2 quantum computing systems with over 4,400 qubits. We're hard at work enhancing Advantage2 with some new capabilities, some new protocols that I'll speak about in a second to really expand the use cases and the ways that we can use our commercial scale and annealing systems to solve problems.

But like Alan spoke about a few minutes ago, we're really focused on driving up scale for our Advantage3 technology. And the way we're doing this is by bringing together multiple superconducting integrated circuits, interconnecting them to form larger qubit fabrics, larger annealing fabrics to unlock larger scale. And larger scale will be able to allow us to address a broader set of use cases.

We're also taking the Quantum Circuits road map. Alan spoke about that a few minutes ago. In the near term, we're really focused on developing and delivering an increasing set of dual rail processors starting this year with a 17-qubit dual rail processor, next year, increasing that to a 49-qubit processor and then the year after that, 181-qubit dual rail processor.

Now the goal with this series of increases in processor sizes to show fundamentally the advantage of the dual rail system in quantum error correction and to allow our customers to start exploring algorithm development with this new capability with the high fidelity, high-speed dual rail systems as soon as we can make them available.

We're also making all of our technology available via our Leap cloud platform. We introduced a quantum classical solver, the Stride solver, and I'll say a little bit more about the Stride solver in a few minutes in 2024, and we've been intensively investing in the capabilities of the Stride solver can support up to 2 enterprise-grade optimization problems with over 2 million variables.

As we develop the dual rail technology, we are making that dual rail technology, first, the simulator and then the dual rail technology itself available in Leap. And then going forward, we're actually seeing quite a bit of interest. And my colleague, Lorenzo Martinelli, will speak a little bit about the on-prem interest in this interest from high-performance computing centers in integrating quantum systems, annealing systems and eventually gate systems into their overall computing infrastructure. And so this is an important part of where we're driving with our technology, what is the best way to integrate our systems, annealing CPUs and our gate model systems into high-performance computing centers.

Our flagship annealing quantum computing product is our Advantage2 annealing system. This is our sixth generation of annealing system product. We focus on 3 key pillars when we were developing Advantage2 to drive performance. The first is qubit connectivity. Qubit connectivity, as you increase it, you're able to actually represent more complex, more complicated optimization problems and map them directly into the fabric of the annealing processor. With Advantage2, we've gone to 20-way connectivity. That means in the fabric of the processor, every one of the qubits is tunably coupled under user control to 20 other devices. This enables you to pose and solve more complex problems with the technology.

The second thing we focused on was qubit coherence. Now coherence is critical for driving faster time to solution. And over the last several years, we've been able to fundamentally make a connection between higher coherence architectures and faster time to solution. We've reengineered our fabrication stack, including materials and processes to produce the Advantage2 integrated circuits, and we're actually seeing more than double the coherence times in the processors. And this is important. This drives dramatically faster time to solution for hard optimization problems.

And finally, we've increased the energy scale. You can think of energy scale as sort of the strength of the interactions that we can actually represent between the qubits or how precise or how high fidelity you can represent these hard optimization problems in the fabric of a processor. Increased energy scale delivers higher quality solutions. So we've pushed hard on all 3 for the Advantage2 technology.

Our Advantage2 systems are accessible today through our quantum Leap cloud platform, and they're available for customers that want on-prem delivery, an on-prem Advantage2 system. And like Alan said, they support enterprise-scale hybrid applications. Our Stride solver is backed by our Advantage2 QPUs and our Stride solver can support enterprise scale optimization problems with up to 2 million variables and constraints.

I want to say a few more words about coherence and the connection between coherence and performance. Coherence in general is what quantum computing technologies are harnessing to deliver computation utility. And as we've driven up the coherence in our processors, we've seen better performance in terms of faster time to solution for optimization problems. I'm illustrating this in the graph on the right, where I'm showing some optimization benchmarking data comparing Advantage2 and Advantage on a fairly standard hard benchmark class of problems that we use internally as a report card or a test of how well the technologies are working.

Shown in the orange are data from Advantage2. Shown in the blue are data from Advantage. And what I'm plotting along the y-axis is relative error. So you can think of this as distance to the optimal solution. You want to be driving down to 0, to 0 relative error. Along the X-axis, I'm plotting anneal time. So how much time you're using per cycle to actually do the annealing process, the calculation.

You can think of this as computational effort, how much time I'm willing to spend to get an answer to the problem. You can see that for Advantage2, compared to Advantage, uniformly across all of the annealing times is outperforming Advantage, often by an order of magnitude or more in terms of the relative error that we can drive down to.

But as importantly, you can see that if you pick a relative error, say, a few parts per hundred and just draw a line across the chart to compare Advantage versus Advantage2, the increases in coherence are not just marginal. They're not just 20% or 30%. But in some cases, for some relative error targets, you're seeing orders of magnitude reduction in time to solution. And this is the power of delivering more coherent processor technology and more coherent qubits in the fabric of our annealing technology.

Alan spoke about our quantum supremacy demonstration. Two years ago now, we posted on to the archive our preprint beyond classical computation and quantum simulation, where we took the Advantage2, in that case, it was even a smaller scale prototype of Advantage2, this was before we released Advantage2, and we studied how well Advantage2 could actually simulate the properties of a magnetic material or separate magnetic materials that were undergoing phase transitions and study their dynamics.

And we worked with a team, an international team of experts and the best classical methods for solving this problem. We actually consumed tens of thousands, I think, up to 100,000 hours on the Frontier supercomputer at Oak Ridge. And we established this classical scaling, how hard this problem was for the best known classical methods. And there's a massive gap between what the classical methods can do and what we can do with our Advantage2 processor.

Alan spoke again a few minutes ago, we can solve this problem in 20 minutes, and we're projecting for the largest problem sizes that we studied would take up to 1 million years on Frontier. We're harnessing quantum coherence to deliver computational advantage. In this case, it is beyond classical. So this material simulation calculation just can't be done with classical computers.

So I want to dig a little bit into annealing's long-term advantage and optimization because you can ask the question, and Alan asked this in his keynote, well, why not use gate model systems, including today's gate model systems to solve optimization problems. Well, there's a growing set of evidence, both empirical and theoretical that suggests that annealing systems will always have a leg up our long-term advantage over gate model systems in solving optimization problems.

In the near term, you can actually just horse race the technologies today against each other. And in fact, there's a large number of empirical studies I'm pointing to at the top of this chart, a set of data from a reset of researchers at Los Alamos National Labs, where they did exactly that. They studied our -- in this case, our Advantage annealing system and compared it against some emerging gate model platforms. And what they found was the Advantage system could solve optimization problems at a larger scale to higher accuracy to lower residual energy closer to the optimum at orders of magnitude faster execution times than the gate model platforms. There is a fundamental gap in problem scale, solution quality and execution time between annealing systems and gate model systems for optimization.

But to be fair, you could say, okay, well, what if gate model systems, as they scale, like if I hand you a scaled up fault-tolerant gate model system, what about that system? How can it compare in solving to -- how can it actually -- how will it do in solving optimization problems? And I want to point to some work that actually came from Google's Quantum applications team, where they actually studied several gate model algorithms, and they did a very, very detailed analysis of the overhead required in running quantum error correction.

And the conclusion was that unless the algorithm that you're running on your fault-tolerant gate model system has a steeper than progratic scaling over classical approaches, all of the benefit from running on that fault tolerant system is erased by the overhead of fault tolerance.

Now it's not to say that there's a lot -- there aren't a lot of algorithms that have very steep advantage over classical approaches, exponential advantage. Optimization is not one of them. And this is some recent benchmarking data that my colleague, Andrew King, who leads our performance research team, ran on our Advantage and Advantage2 systems on a slightly different flavor of optimization. In this case, it was a response to an initial paper from IBM on multi-objective optimization. Here, it's a more complicated optimization task. You're actually given a range of objective functions and you're asked to optimize over that portfolio of objective functions.

What I'm showing in the plot here are in the blue and the orange are data from our Advantage2 and Advantage system on how well we do at that portfolio of optimization, that multi-objective optimization against the red, which is empirical data from the IBM system. And you can see that we're orders of magnitude, up to 1,000x faster in terms of that performance differential, in terms of the execution time for this hard multi-objective optimization function.

Also interestingly, in the green, it shows a perfect noise-free simulator of how well that gate model system could do. And even in that case, we're still opening up the gap and have opened up the gap by orders of magnitude. This is just more evidence that annealing systems have this long-term advantage in hard optimization problems.

Alan spoke about our annealing scaling road map that we're incredibly excited about. With the Advantage2 system that we launched in 2025, the fabric of our processor at the core is a single superconducting integrated circuit that contains just over 4,400 qubits. We're hard at work on packaging technologies, superconducting interconnects that would allow us to take multiple superconducting integrated circuits and assemble them into a larger processor fabric and tile out this processor technology to grow the scale of the systems.

Why do we want to do this? Well, as we grow the number of qubits, as we grow the qubit count, we expect to be able to unlock broader utility in quantum optimization along with a lot of other use cases. And so we're quite excited about scaling the technology up to, first, 20,000 qubits, with the eventual goal of hitting 100,000 qubits. But the other piece of this is scalable control.

We want to do this without blowing up the number of control lines that go into our cryogenic enclosure. So like Alan said, we have just recently completed a design for some prototype scalable I/O that should bend and reduce the amount of lines that we need to control our devices by even more than we actually have right now. And this really should -- our goal is to get to this technology node without substantially growing the number of lines that need to go into our cryogenic enclosure.

I spoke a little bit about some work we're doing with Advantage2 on new protocols, broadly unlocking what we're calling analog digital quantum computing, but specifically capabilities, new capabilities that we're making available in our commercial scale annealing systems. And again, specifically, the ability to target per qubit in situ excitations in the fabric of the processor and then the ability to read out the state of the qubit in an arbitrary basis.

Why do we think that these 2 capabilities are important? Why are we actually putting some resources into this development? Well, fundamentally, we think that these 2 capabilities, and we've actually seen this with some initial results, should unlock a major expansion in our ability to do magnetic material simulation. So we're incredibly excited about these tools to basically unlock our ability to simulate the behavior of the magnetic materials at even larger scale and more complicated lattices.

But looking forward, the ability to read out in an arbitrary basis will allow us to pull much more complex rich distributions from our annealing GPUs. And we have some great ideas on how this could actually really drive more utility in both machine learning and blockchain applications. And I'll speak about these 2 exciting emerging application areas in a minute.

But I want to highlight some of the data that I and the team are particularly proud of, which is basically a demonstration of this excitation and these readout protocols within the fabric of our Advantage2 systems, where, in this case, in this plot, we've arranged a subset of qubits in the Advantage2 processor, a 1D chain in a line, 120 qubits, have a uniform coupling between these qubits, and then we deliver an excitation pulse to the middle of the chain, and then we watch that pulse coherently propagate both in time and space across the chain and over time.

You can see here, Qubit index on the plot really shows where in the chain you're seeing that excitation and along the Y-axis, and along the X-axis is basically a delay. You can think of this as an evolution time. So you can see these coherent oscillations. This excitation get delivered to the center of the chain and then propagate coherently exactly like you'd expect from theoretical predictions.

I want to speak a little bit about our Leap cloud platform. We're incredibly proud of the Leap cloud platform. We launched it in 2018, and it's a means by which our customers access all of our technology, both our hybrid solvers and human CPUs and very shortly, our dural rail technology. We make available to our customers a set of tools in the Ocean software development kit. This is an open source software kit written in Python, and it's meant to make it extremely easy for formulating hard optimization problems for our technology. And we have a growing set of tools for researchers that are really interested in using our technology to explore new use cases and new algorithms.

The Leap cloud platform has high-performance, high-availability APIs, the suite of quantum classical hybrid solvers and a dashboard that allows you to kind of monitor your usage and really keep track of how you're using the Leap cloud platform and the solvers that you have available to you. And at the core, we have fast fiber connection from the Leap cloud platform to each one of our annealing systems. Both Advantage2 and Advantage are available. We have 4 production systems online available through our Leap platform today.

But this isn't a research platform. This is engineered to support business, to support enterprise-grade operation. This is a platform that is really singular in the industry in terms of its performance. We've engineered to be high availability. We want the solvers to be available to enterprise to business to solve their problems at any time. We don't want the solvers to come in and out. This is high availability, and we've had over 99.9% availability since we launched Leap in 2018.

It's a real-time platform. There's no extensive queuing. There's no need to wait. There's no need to reserve your time ahead of time. You submit a problem to our Leap cloud platform, you get an answer back in subsecond response times. We have SOC2 Type 2 compliance to give our enterprise customers the assurance that we have high security and that they can trust the platform to be there and to be good stewards of their data.

And finally, we're actually offering -- we're putting our money where our mouth is. We're offering service level agreements to our enterprise customers to basically guarantee the uptime, the availability of the platform for their operations. And I want to say this is, again, differentiated, and I think gives us a very strong advantage in the quantum computing marketplace. And we've, from the ground up, engineered this to support business.

One of the flagship solvers that is currently available in the Leap Cloud platform is a Stride hybrid solver. We launched Stride in 2024. And Stride at the core is combining classical and quantum resources together to solve enterprise-grade optimization problems. Stride can support up to 2 million variables and constraints and is really starting to deliver differentiated performance on a growing wide class of use cases. Something that we're particularly excited about that we've introduced into the Stride solver several months ago is surrogate modeling, the ability to import a machine learning model directly into your optimization workflow.

So why is this important? Well, going back to one of the examples that Alan spoke about where you may want to use a machine learning model to predict demand if you're a retailer, and then optimize that demand using our optimization solver or engines. And it's sometimes very hard to come up with a closed mathematical form for that data set for that demand. So what you can do is train what we call a surrogate model, machine learning model that can actually serve as a model for that demand and then embed it directly into your optimization workflow. So we're actually very excited because this is really enhancing the behavior and the ability of the Stride solver to open up a growing set of use cases.

And one particular example of a hard optimization problem that we benchmarked the Stride solver against, I'm showing in the graph on the right is a problem called satellite placement. And this is actually a fairly hard problem because there's a discrete components like where are you putting satellites in a particular orbit in terms of an ordering operation. And there's also a continuous component. How do we actually do the fine-tuning of where those satellites go with the overall objective of minimizing my gap coverage.

I want to actually have no gaps in my satellite coverage in terms of the ground coverage. So what we're showing is benchmarking data from the Stride solver along with several other solvers, traditional solvers for solving this problem. We're on the Y-axis and plotting median gap. So you can think of this as distance from optimal, you want to drive down to 0 gap as a function of the number of satellites in your constellation.

So different -- all the different run times and comparisons for this problem class, the Stride solver is getting very close to or is getting the optimum gap driving to 0 gap and is outperforming the other solvers.

I want to switch gears in the next minute or 2 and talk about 2 emerging application areas that Alan spoke a little bit about in his keynote, quantum AI and blockchain. In quantum AI, one of the things that we're particularly excited about is directly using our annealing CPUs to enhance and to improve machine learning models. Now at a high level, you can think of a machine learning model as the training of a model and construction of the model as an exercise in compression. You've got a very large data set. You want to learn the structure of that data set, you want to lower the dimensionality and come up with as compact a model as you can to represent your data and then use that model to generate new high-quality versions of the data.

Our vision for how to integrate our annealing CPUs directly into this process is to go through an encoding process where we take our initial data set and we use a classical neural net and encoder that then map that data set into a compact lower dimensional latent space. And in that latent space, our annealing CPU learns the best structure in that latent space to compress that data and to represent that data in as few variables as possible. And at the output, you basically take samples or new samples from that latent space to generate new versions of your data.

Now the critical thing here is that you want to train the QPU simultaneously with your encoder and decoder neural nets. So when you do that training simultaneously, the QPU is able to learn the best distribution, the best sort of shape of that latent space to compress your data to high quality and deliver a particular sort of objective function in terms of what you're trying to train towards. So then when you're actually prompting the QPU or prompting your model for new versions of the data, you're sampling from the QPU and then decoding new versions of your image and new versions of your data set.

So where do we think that there's long-term value in this overall structure? Well, the first is that the QPU is incredibly good at learning complex distributions and the QPUs are incredibly good at generating high-quality samples quickly. And in fact, there's some growing evidence that sampling -- that core sampling capability, generating novel samples from a learn distribution, CPUs have an advantage of our classical approaches. And so we're harnessing that directly and wiring that into the generative AI models.

And a great example is what Alan spoke about a few minutes ago. This is a drug discovery application where the data set instead of images is a molecular data set. And we use the QPU to learn the structure of this molecular data set. And we saw -- we were able to compare this quantum classical data set to a classical-only version of the model.

And what we found is across 3 different metrics, the quantum classical model outperformed once trained, outperformed the classical model, both in the metric of molecular weight. It was able to deliver high-quality samples that were smaller and therefore, more valuable because they're easier to test and develop, a metric called lipophilicity, which is a measurement of sort of how sort of fat soluble the molecule is, and we could get much better lipophilicities with the quantum classical model.

And then finally, something called drug likeness. The structures were sort of best suited for kind of human drug candidates. Across all 3 of these metrics, the quantum classical hybrid model outperformed the classical model. Again, the key view here is providing discrete samples from a complex distribution and then those samples are translating to better outcomes, to better drug candidates.

The second emerging application area that we're quite excited about is Quantum blockchain. So Alan spoke about post-quant labs. This is a partner that we're working with on a quantum classical blockchain and the testnet is currently operating. And actually, there's been an immense amount of interest across the globe. This is kind of a map of the different participants and the kind of the compute nodes that are currently online as part of this testnet. There's over 18,000 users of the testnet and over 1,600 compute nodes currently running in testnet phase. And when the QPU is on, our Advantage2 CPUs are always available in Leap, but we're making them available for the testnet up to 5 minutes per day. And when those CPUs are on, they're completely dominating the mining rewards. So like Alan said, we're moving to a phase where we do a detailed benchmarking study where we really show the performance of the QPU as is competing with classical miners, a set of GPU and CPU miners to win token rewards.

But longer term, we're very excited about fundamentally a quantum-only proof of work. We published a research paper about 2 or 3 months ago, a peer-reviewed paper that demonstrated both a proposal for quantum proof of work along with the demonstration using 4 production Advantage and Advantage2 systems that were online distributed through North America. And we showed that we could run a stable blockchain with a quantum proof of work at the core indefinitely using these annealing GPUs.

Why is this important? Well, from the ground up, this is a quantum-safe protocol because the fundamental proof of work for this blockchain is quantum, there's no way it can be spoofed classically. So it's quantum safe by construction. The other thing that we're quite excited about is that we've estimated that blockchain supported by this particular quantum proof of work could be 1,000x more energy efficient, orders of magnitude more efficient than a classical analog of a quantum proof of work blockchain. So fundamentally, you have a potential for a quantum-safe technology that is substantially more energy efficient.

I'm going to switch gears now and spend a couple of minutes kind of talking a little bit about our gate model program. And then I want to invite my colleague, Rob, up to the stage, and he's going to do a deep dive on the dual rail technology and where we're going with the gate model program.

So why are we developing gate model systems? Alan did a fantastic job of motivating it. We really see annealing and gate systems as complementary. We want to address the full addressable market, the full TAM for quantum computing technology. We see use cases that are complementary to the annealing systems, and I'll speak a little bit about which use cases we see as really emerging as important in the near term for gate model technologies. And we have a road map for delivering 100 logical error-corrected qubits. And we think that this will unlock early commercial utilities with gate model systems.

So why now? Well, the merger with Quantum Circuits is bringing a leading superconducting qubit technology to D-Wave. We have cryogenic control that allows us to scale superconducting systems. We can use our 15-plus years of experience with engineering this control and apply it to the dual rail technology. And the combination of these 2, we think, will allow us to win the race to be the first to deliver fault-tolerant systems to the market.

Alan spoke a little bit about the TAM. I won't get into the numbers in detail, but I just want to double-click into some of the use cases that we're most excited about. I think some of the likely earliest use cases are broadly in the novel drug discovery, novel materials discovery domain. So this is broadly quantum chemistry, pharmaceuticals, trying to build your molecules and measure them ahead of time with your gate model systems before building in the lab. We see a massive opportunity in the near term. This is a huge space and is a big segment of that TAM that Alan spoke about. We see value beyond our annealing system for gate systems as a scale and really doing heavy lifting for machine learning use cases.

There's a broad set of use cases in the finance industry on sort of solving hard differential equations to drive sort of better performance for portfolios, better trading outcomes. We see this as probably needing larger systems, so maybe later-stage use cases, but an incredibly large market that we've identified.

And finally, there's a space of cryptography. So this algorithm really was the algorithm that helped launch the birth of the quantum computing industry. Gate model systems are going to be able to start attacking encryption as they scale out. All of these use cases here are motivating and the breadth of them are really motivating our investment in building gate model quantum computing technology.

So with all the pieces that we have in front of us, we're ready to scale and win the race. We have an industry-leading qubit technology that Rob will speak a little bit more about. This technology and its current performance gives us an achievable path to fault-tolerance systems with an order of magnitude fewer physical qubits required to encode our logical qubits and get to fault tolerance.

And finally, our experience with cryogenic multiplexing, our scalable manufacturing and our experience with cryogenic -- reliable cryogenic platforms is opening up the path to a commercial gate model technology that's reliable, that's production grade that we're excited to put in the hands of our customers as soon as we can.

With that, I want to welcome Rob to the stage, and he's going to do a deep dive on where we're going with the gate model technology.

Thanks, Trevor. Yes. So hello. It's nice to be here this afternoon. Thank you all for coming and for your time. So yes, I'm going to begin here by describing in simple terms, what is a dual rail qubit and why is it so different?

So the first thing you need to know is that it's a completely different paradigm. It's something we call an erasure qubit, which has the ability to detect errors kind of built into the hardware. And it's shown in kind of cartoon form here. It's a composite object that's composed of a few simple parts, some of them being the Transmons, which we actually developed in our lab maybe 20 years ago at Yale that or what everybody else uses, but they're really just used here for the input and output. It's to create excitations in the heart of the system, which are a pair of microwave cavities. And what we do is we store a single microwave excitation in those cavities and it's the stability and long lifetime of those cavities, which make this the most coherent superconducting qubit that's out there in the marketplace.

So let's talk a little bit about how in the dual rail, we store information and process it. So the logical states of this dual rail qubit are those in which a single microwave photon excitation that sort of cellphone frequencies of the electromagnetic field is shared between these 2 microwave cavities. And you can have the photon existing in the left cavity, let's say, call that logical 0, or it could be in the right cavity, call that logical 1, and then we can manipulate that information by sending an electrical signal to another superconducting device called a coupler that allows us to exchange this exitation without losing it between the 2 cavities. That also allows us to create states like a super position, which is 0 plus 1, where the photon is actually shared at the same time between these 2 superconducting boxes or cavities.

Now what's special about this dual rail and this erasure qubit is that 90% of the errors, the dominant errors are just those in which after a sort of a millisecond or so, the exitation is lost. And that leaves us in this unique state where both of the cavities are empty. And so I like to say sometimes that the dual rail has a 0 in a 1 like regular qubits, but it also has this third state, which basically flags that there's -- it's invalid. There's been an error, and you can use that to either boost the performance by rejecting those shots or to give you extra information that dramatically aids in the full task of error correction and making that more efficient.

And so this ability to detect the errors is really one of the key capacities. It's also something which is interesting for algorithm developers and designers because it gives you more information than you get on a traditional kind of machine.

We've talked about how this dual rail is really unique in the space in its performance characteristics. And I already mentioned that we can detect 90% of the errors and boost the fidelity. So we can have 2-qubit operations with something exceeding 39s, 99.9% fidelity. That means an error only occurs about 1 in every 1,000 operations. And that's the kind of fidelity that you usually don't associate with superconducting platforms.

And so this chart sort of shows 2 of the key metrics, right? One is the fidelity on the vertical axis and the gate speed on the horizontal axis. And in the quantum space, very broadly, there are kind of 2 main types of qubits. There are trapped ions and neutral atoms, I call them the God-given qubits, the sort of natural qubits they will tell you, microscopic objects that usually have higher fidelity, but are also associated with much slower operation times on the order of milliseconds.

The other category is solid state, mostly superconducting qubits, which are very fast, but are usually noisier and more error prone. So they're kind of on the lower right. And with the dual rail and our ability to detect the errors, we retain the manufacturability, the scalability and the very high speed, microseconds rather than milliseconds of superconducting systems, but we're able to boost the fidelity and the performance and get sort of the best of both worlds.

Okay. So where are we today in gate model computing, and where are we going? So to unlock the full potential, as we've already discussed this afternoon of gate model computing, you really want to be on the right in this fault-tolerant era. In that realm, you are using something called logical qubits to dramatically suppress the errors of the computer below those of the physical pieces to extend operations and make much more complicated and more powerful computations. Maybe you want to do 1 million computations, which is 3 orders of magnitude lower error rates than you can get with physical qubits today.

Now most of the industry is in what we call the NISQ era on the other side of the chart, NISQ, standing for noisy intermediate scale quantum. Here, you just have physical qubits. You can do maybe tens or hundreds of operations and the use cases and the applications are somewhat limited. At D-Wave, we're the first to be in what I like to call the post-NISQ era. So we have this error aware feature of our machine. The NISQ machines just give you out a result. It's mostly noise, and your job is to try and figure out what the quantum computer or the quantum algorithm was trying to tell you.

But with our error detection, you can either reject shots that have noise and keep only the pure shots or you can learn how to do other sorts of things. Now an important point here is that really, this is learning how to program the quantum computers of the future. Even when we're in the fault-tolerant regime, we're going to have to deal with errors. You're always going to want to run a computation a little bit longer than you can, given your amount of error correction. And so you will want to use the information of error detection on top of error correction. But in this post-NISQ era, we're also very excited because now we really can attack the problem of error correction and doing so efficiently to get us into the fault tolerant era.

Okay. I want to do my version of explaining this chart, which Alan already presented and explain for you again kind of the general way that error correction works and what efficiency means in that context. So the general idea of error correction is that you will construct out of some collection of physical qubits, the number of which is sort of the horizontal axis on the bottom, you will construct a logical qubit that redundantly encodes one usable bit of quantum computable information in that distributed among that collection of physical qubits.

And the idea is that the error rate now shown as the sort of difference from one, right? So lower error rates here being smaller and smaller errors or longer and longer computations is shown in a logarithmic scale on the vertical axis. And if, and only if, your performance of the elements is good enough, then the idea is that you can increase the amount of redundancy, you increase the number of physical qubits making up your logical qubit or you increase what's called the distance of the code. Those are the numbers going across the top 3, 5, 7, 9. And for each increment in the code, you get another multiplicative factor that reduces the errors. This factor is the number called lambda or the error reduction rate.

And in conventional superconducting systems, they've been working on error correction for 15 years or so, and they've been able to get now finally to a lambda of about 2, meaning that you halve the error rate. But with the performance we've observed already in our prototype systems with the dual rail qubits, we anticipate that this lambda factor can be as high as 10. And that means that every time I increment the code, I'm going to reduce the errors by another order of magnitude. And so let's say, I want to get to a part per million, I want to run a computation that's 1 million steps long. On a conventional machine, which is this upper graph, you'd have to go way out to the right with a very large distance code and a massive number of physical qubits making up your logical qubit, and you have to wait for that scaling to happen. But on the other hand, with the dual rail, you can get down to parts per million with maybe only a few hundred qubits. And the near-term part of our road map is we're building these systems to do distance 3, 5, 7 and 9 and to demonstrate this high value for lambda and suppression of errors down to this kind of hopefully part per million level.

Okay. So that's something that now makes sense to scale in gate model. And what do you need to do in order to be able to carry out that scaling? Well, one of the challenges is for a still large number of qubits for a usable computer, you have a large number of signals and wires that need to be brought down into the cryogenic environment.

And so a solution for that is to use some amount of cryogenic control and logic and to be able to sort of multiplex and bootstrap your way to using one wire to control many qubits. And D-Wave knows how to do this. They've done this in terms of scaling the annealing processors to this point. And so we're very excited about being able to now combine the technology of the dual rail that we had developed at Quantum Circuits with this knowledge base and intellectual property that D-Wave has. And so that's the other ingredient in scaling efficiently.

Another important aspect of scaling is to have systems which are mass producible and reliable and manufacturable. And the dual rail, although it's a different concept in superconducting qubits, it uses very much the same kinds of materials and fabrication processes. And so we're able to adapt the -- some of the techniques from the semiconductor industry and various foundries in order to be able to engineer these systems and make more larger scale and more useful machines.

This is, again, repeating a kind of table that we had before, emphasizing kind of speed and performance or -- but for logical qubits. So we want to move the industry to starting to talk about what your error correction really does and what's the performance of your logical qubits. And so you still care about speed. Now in your error-corrected machine, it's the time it takes you to measure one round of error correction that is basically the basic clock speed, and that's a small multiple of the individual gate speed, it's less than 5 microseconds for us and for superconductors. And for trapped ions and neutral atoms, it's many orders of magnitude longer, milliseconds or even more in some cases.

And again, the state-of-the-art is really only to be able to produce marginal gains in your error correction, not these lambda factors of 10. And so that's the second column, this important error reduction rate, which tells you, again, how fast you're able to suppress the errors or how you're able to make a high-performance logical qubit with over an order of magnitude fewer actual physical elements.

So we're very excited to be talking about our gate model road map. And it's divided here into 2 eras. The near-term era, this post-NISQ era that I mentioned, we are delivering over the next 2.5 years or so a sequence of these dual rail processors that are based on our existing prototypes and deliver the first error correction layered on top of the error detection that's built into the qubits.

And as we go from the 17, which can do a distance 3 code to the 49, which does distance 5, we expect to see one extra multiplicative factor of lambda in the error reduction factor. So this 20x is the net error reduction. And then with the 181, we can do distances 7 and 9 and get 2 more of those factors to get to error suppressions that are overall more than a factor of 1,000, we hope. And so that is the near-term road map.

And then really, when we have this proof in hand of the error correction efficiency, we really have a blueprint for how to scale into this fault tolerant era rapidly and efficiently. And so we're targeting in 2030 systems with of order 1,000 physical qubits which would house 10 logical qubits, but at error rates of perhaps a part in 10 to the 6, allowing millions of computations on those qubits. And then with just another factor of 10 increase in the number of elements to get to 100 logical qubits with millions of gates, which we anticipate could bring us to the sort of initial quantum utility for these systems.

Just drilling down a little bit on this near-term road map. I've already mentioned how the 17, 49 and 181 are very useful proof points for error correction. But also, they allow us to begin having people use these systems and learn how to program in these new error-aware ways. So they're dual-use. And so you can use it as 49 physical qubits, but with the boosted fidelity that you get that's beyond what ordinary superconducting qubits can yield.

Okay. So that's what I wanted to tell you about. I think with this acquisition of Quantum Circuits and joining forces with D-Wave, we're really in a very interesting place where we're ready to scale and win the race to fault tolerance. So we have the most performant, highest coherence and best qubit out there, the dual rail. It has these unique capabilities that allow you to do error correction much more efficiently. That extra information, the higher performance allows you to deliver logical qubits with over an order of magnitude fewer elements. And then by utilizing the already developed techniques for multiplexing robust cryogenic systems and mass-produced manufacturability in foundries, we're really excited about being able to bring to market these series of machines.

So with that, I think Trevor is going to join me, and then we'll take a few Q&A after one more slide.

Thanks, Rob So, again, we are focused on a dual platform strategy because we see annealing and gate systems as complementary. Rob talked about our path to the fault tolerant era and the path actually unlocking utility with gate model systems and bringing to the stage where they're commercial alongside our current commercial annealing technology.

So with that, I think we have a few minutes for Q&A, I think. Thanks. So John?

Trevor, I appreciate you putting that slide up last. I had a question about that one before. The either/or annealing versus gate model, it's very interesting. I really haven't seen you talk about that in detail before. How do you delineate those problems, those very important problems, machine learning, cryptography, drug discovery about which ones D-Wave might be able to handle versus gate model, the size of the problem? Is it the system road maps? Because that could bring that -- those opportunities 5 years forward versus waiting for gate model. So I'm curious about that.

Yes. I mean that's a good question. There is that overlap in the TAM, like when Alan showed the TAM slide, there are some of those areas where we expect annealing and gate systems to address and broadly in the area of cryptography and linear algebra. In fact, I spoke about sort of what our near-term plans are in quantum AI, where we're really using our annealing systems as critically important sampling engines to make machine learning models potentially smaller, more efficient, make the training faster and to make inference faster.

And so we actually see real near-term wins with annealing architectures. But there are some algorithms that are fairly exciting, including one that was proposed by the -- again, the Google Quantum Applications team, where they show that once you have error-corrected logical qubits with a qubit count in the 50 or 60 there's actually a signal processing algorithm you can run to really allow to compact data sets for machine learning applications. And I can forward you the link to that paper. So that's an example of a machine learning use case that is emerging as important that could be enabled by gate systems. So the short answer is I think we're going to learn a lot over the next several years, but we're actually attacking some of these use cases with our annealing systems today.

Konrad, Morgan Stanley. Thank you for these very informative presentations. Really an infrastructure-related question. Can you address the cryogenic environment? Because the cooling systems operating at absolute Kelvin like outer space at all really hits the bottom line. And as much as you can get energy efficiency through rapidity, that is also a critical factor. So just a little bit of color on that subject.

Yes. I can take it and you can add commentary. So cryogen -- it's true, we need to cool our technologies, both gate and the annealing are superconducting technologies that need to be at millikelvin temperatures to be in cryogenic enclosures. But I see cryogenics as a solved commercial problem. We buy off-the-shelf fridges from vendors for both our annealing and our gate development. Those fridges are high reliability. We've actually spent at D-Wave 15-plus years engineering out all of the issues that happen when you try to run these systems for years at a time. And we've had systems in customer data centers for almost 5 years running continuously with no interruption.

Our current systems, we are drawing about 12 kilowatts of power. And that for our annealing systems, we've had the same cryogenic enclosure for 6 generations of annealing system. And so that overall 12 kilowatts of power is a fixed power cost over multiple generations. So it is true that they do consume power, but I really don't see this as a cost driver for the technology. And in terms of overall complexity, I see this as an off-the-shelf technology that is -- basically is commercial today.

So Trevor, what do you think we'll need for the 100,000 qubit system?

A slightly bigger fridge, but not much bigger. Still something that we can build off the shelf -- we can pull off the shelf.

Good point, not much bigger.

And what you need is a different size and shape of the enclosure, but not necessarily more refrigeration. So you can still adapt kind of the conventional machines that are out there.

I mean I think these comments that superconducting requires a football field size system and $1 billion is nonsense. That's part of the hype. It's just nonsense.

Just a quick question on -- there are a ton of road maps out there, a lot of different companies with big goals. I wanted to just get your thoughts on -- and it was great to see you put out the gate-based road map this morning. But in a scenario where the other people do what they say they're going to do. What would be your plan to compete with them, let's say, in a situation where someone else had a really fault-tolerant computer in the market before yours, and you have the leverage with the annealing side, but how would you plan to compete then?

That's a great question. What Rob showed in terms of our road map to fault tolerance, I consider a credible, achievable road map but still aggressive. So there's still a lot that we are doing to integrate the multiplexing, the cryogenic control, and we're needing to scale up the technology. So I see this as a credible, achievable but aggressive road map. I see some of the other road maps as I'd be very excited if some of the other efforts actually nailed exactly what they're saying on those time lines. But I think we are really focused on how to actually do this properly and achieving this in a realistic way.

So I want to say just a little bit more. Okay, let's suppose for a minute. Let's suppose for a minute that we get a scaled error corrected neutral atom or trapped ion system 2 years ahead of us. Okay. And then 2 years later, we have a system that's 1,000x faster. I don't care about the 2 years, right? I mean, at 1,000x faster, they're out of business. And let me just kind of give you a data point to hang your hat on.

I think we're all kind of intrigued, maybe even excited when a few weeks ago, Google said, "Hey, we can break the Bitcoin protocol with 500,000 superconducting qubits and we can perform the computation in 9 minutes." They actually didn't say we can perform the computation in 9 minutes. They just said we can break it with 500,000 qubits. Then John Presco from Caltech and [indiscernible] came out and said, "Well, with neutral atoms, we can do it with only 10,000 qubits." Okay. Good. 10,000, a lot less than 500,000.

Go read the paper. How long will it take to perform that computation with those 10,000 qubits? Months, not 9 minutes, months, okay? That doesn't break the protocol, right? Because you you're not going to win a block if it takes you months to perform the computation. Google at 500,000 qubits would win it. So the point is once we get superconducting anything, we think it will be us because we have that error correction advantage, the ability to error correct with far fewer logical qubits. But once we have superconducting, it's game over for everybody else. Nobody is going to want a slow quantum computer. It's just physics.

And the point is we scale faster. Our capabilities will go faster as the number of qubits increases. So that also helps.

Quinn Bolton with Needham. Trevor, I just wanted to follow up on the machine learning example you gave where you mapped to the latent space. How broadly applicable is that? The work that Shionogi did seems like it's given them an advantage. It certainly seems like it could give others advantage. So are there limitations to that approach? Or do you see pretty wide interest in...

Yes, that's a great question. We're actually doing some internal development on a diffusion-based model. So where the latent space is able to act as a prior for image generation for image data sets. But we chose that basically because there's a lot of image data that is kind of publicly available and a lot of benchmarking results out there. So it gives us basically a really good sandbox for doing model development.

But I'd say that this general idea of compression using the QPU to learn the structure of the latent space and then efficiently sample is very generally applicable. We're just choosing to focus on, for example, the customer project, they brought us some molecular data set, and we're actually using public image data sets to actually do the model development on. But we see this overall protocol, this overall approach of enhancing a generative model as very generally applicable. And we're actually getting quite a bit of -- a growing set of interest from customers.

In fact, we've had a more research-focused customer, high energy physics customer use exactly this type of technique for generating sort of simulated data for high energy physical experiment. Again, they're seeing a real computational challenge in terms of Monte Carlo simulation of next-generation experiments in CERN in Switzerland. And so they have a quantum-assisted AI model that they've built using a very similar structure, and they're seeing up to a 10x faster sort of inference and sampling capability as compared to GPUs. So I think it's extremely and broadly applicable.

And we have two questions from the virtual audience. Kevin?

The first one is around manufacturing and particularly related to the announcement a couple of weeks ago from the government about funding for Global Foundries and IBM's Anderon. Does that change D-Wave's manufacturing strategy at all?

So I think our strategy from the beginning when we first started developing our Annealing Systems was to harness commercial foundries to build our technology and adapt those processes to build our systems. And so that continues to be our strategy. I mean we are looking and open to new foundry capabilities as they come online, including new foundry capabilities that could come out of the kind of the portfolio of efforts that the CHIPS office recently announced funding for. And that continues to be our strategy. We want to actually use like commercial foundry services and adapt them to produce our technology.

Okay. And the second one is, I guess, I'll call it the user experience. We spoke about solving problems that are impossible to solve on a classical computer. So the question is, what do users do today? If it's impossible, they must be doing something. So how are they solving problems? And then do we expect any closing of the gap from the -- or response from the classical computer companies?

So I mean, Alan spoke about sort of companies that are getting ROI from using our technology to solve their optimization problems and running their production operations. And so those are the class of customers that are actually accessing and benefiting from our technology today. I mean there are always advances in classical computing methods. So -- and we'll look carefully at all of those advances as the field progresses.

But the gap that we're opening up means fundamentally, quantum computing technology, both annealing and gate systems are harnessing quantum mechanics to solve problems that just can't be solved classically at scale. And so unless classical computers can figure out how to harness quantum resources, they become quantum computing technologies, we think that we will continue to open up a gap of classical approaches for a lot of these hard problems.

So maybe just say a little bit more.

Yes, a lot of these hard optimization problems, companies are solving today. But what they're doing is trying to simplify the problem and use heuristics to come up with what they hope are good enough solutions. The quantum computer is able to solve the complete or more of the problem faster with better solutions. So even if we're not talking about problems that can't be solved classically, they can be solved just not to optimality, there's still a significant benefit for the quantum computer.

Okay. With that, we're going to wrap it up and get back on schedule here. Thank you.

Please welcome to the stage Chief Revenue Officer, Lorenzo Martinelli.

I'm shifting gear here, okay, from all these technologies. Hopefully, you know all about it. Bid you can't go home until you fill in the test and you can pass everything was presented. But I'm going to talk to you for only 10 minutes on the commercial side.

I've been here as Chief Revenue Officer for 2 years. So I'll tell you what we've done and address one of the questions that came up is how do we get more people to take advantage of it. And to be honest with you, I see this one event. You're my evangelist. So take what you have, and we'll tell you, and hopefully, you help us get more people to ask a question that we can address. So my team is a global team that helps customers like AT&T figure out how to use our quantum computer successfully. And what do they care about it? Well, not that many people care about the quantum, all the questions have been covered here. What they just care about is I got a complex problem. It takes a long time on classical or I can only have a simplified version of that problem solved.

And just my background, I'm an engineer. I'm an optimization for 25 years. What you do, you simplify the problem so you can solve it. That means it's not an optimal answer. It's not the best answer. So I go look for people that have a problem, they know quite well that is either a simplified problem, so they don't have the best answer or it takes too long and see if we can do a better job. And then you got people like AT&T. The problem is people like me do not know that quantum exists. When the recruiter tell me 2 years ago to D-Wave, I said, that stuff is 20 years away. Part of the confusion this gate versus annealing. People do not know about an annealing because we are people that offer an annealing. So what's the best proof that we're getting some traction?

Well, over the last 18 months, we've had 26 customers just like AT&T publicly say, "Hey, this thing is actually real. I must be solving problems today." You can find this in our press releases. You can find people that presented our events or industry events, an event like today. And of those 26, the one I'm most focused on is what I call Lighthouses. Lighthouses are customers that are leader in the industry and they get the attention of their peers, like AT&T, for example. This is a list of 15 of them that in the last 18 months went out and said, "Wow, I solved the real problem, I get a better answer and get it faster."

For example, in conversation came up, Andrew, when Andrew spoke, in their industry, people paid attention. When BASF, which is the leader in chemical, issued a press release with us, people pay attention. Shionogi, same thing in their industry. And of course, some of those are also big global brands that we all know about it. So my team's whole job is to go after these kind of companies and ask them what are your complex problems and then see if we can provide you a better answer, a faster answer. And we now have a very well-proven method that I'm going to go through here to basically go from identify the problems. We go through a discovery process, but we figure out which problems do you have today. So this is not an experimentation.

Don't come to me with a problem that you can do on classical and you just want to try on quantum, because it makes no value at that point to try. If you can solve it on classical, you solve just as one quantum. It's like having -- I have a bicycle to go to get my milk at the corner store, somebody comes to me with Formula One car, I'm not going to get that any faster, okay? So I'm looking for problems that are that complicated. Usually, people come to us with that. There are people like AT&T that knows exactly what those problems are in their business. And we go through and take a look at it and make sure that the mathematics of that problem is good fit for our systems.

The second thing we always ask is the data available because a lot of the time, we're talking about data sets that they don't have because classical can't do it. And then we prioritize. Let's start with one. It's critical to find the landing use case that is just a killer, with home run. And then we go through what we call a proof of technology. We want to prove to you with your data set because they don't care about generic benchmark. Let's run your data set for that use case, and I can tell you exactly how much better is the answer for you and how much faster is it with your data. If we can do that, great. Otherwise, let's go on to the next problem.

Once we know that it is the right problem, then you got to turn it into a formulation that can work for all your permutation combinations. So we call a proof of concept. We need many data sets. And we're talking about maybe a couple of months in proof of technology, proof of concept, maybe same time frame. By the end of gate 2, the customer has a formulation that is production ready. Of course, that's only the calculation side. They have to do the work to get the pipeline of data, which usually is more data into our systems. And then once we provide an answer quickly, feed it back into the process. So for the first use case that we tend to look for, hey, one way, you already have a pipeline ready, you know you can just feed into your existing, so there's a minimal amount of work to do that.

And then we want to test it out in production, but limited just to compare, is this giving us results and also to prove out that the value we expected in dollar terms actually matches expectations. And then we go full blast production. We have customers that have been running for 3 years in production. We have systems that can handle that, no problem. The whole process here, maybe 6 months to 9 months to 12 months, depending, not so much how dependent, but how quickly can they get the data set ready and they're ready to go. For example, the customer that Alan was talking about that went through the whole process of Fortune 100 companies in a bit less than a year. And the major bottleneck was waiting for to get some of the data pipeline ready for full production.

Now here are some of the example people that have talked about their results publicly. Interesting enough, if you get into some really interesting use cases, they're not so keen on sharing that they are achieving that because they don't want their competitors to know. So getting them to speak is a little hard. But Andrew, for example, talked about it, it's a defense type application, threat mitigation type of effort. You imagine you have a whole bunch of inbounds and you got to figure out how to take them down. And we could come up with an answer that it's 10x faster than what they could do with their classical systems and about 9% to 12% better hits. That's meaningful in their industry.

To me, it doesn't mean very much, but they got super excited. And you can hear excited there in the video on YouTube, if you want to listen to it. BASF is a chemical environment, so it's process manufacturing, filling out containers and so on and so forth, reducing the scheduling time from 10 hours to just seconds. Imagine your demand changes, you go to make a decision if you have to wait 10 hours. A better example that we're working on here that is not listed here is how many of your people have flown here or you all local. Well, how many times you get stuck on a plane that maybe there's a weather issue, equipment issue and there is some change, recovery disruption. Well, that's an interesting problem that I'm very excited about it to fly a lot. And it's not feasible in classical computing to solve that problem.

So we're aggressively going after that. If you can solve that problem in less than 20 minutes, it would be great. And I think we're getting pretty close there. NTT DOCOMO, it's one that published results. The real problem they wanted to solve takes about 1,000 variables and constraints. You can see the paper. They can't do that in classical. They had to simplify it down to about 13% of those, 130 constraints, and it takes them 26 hours to run. They're now down to -- first of all, they want to figure out if I can do 1,000, how much better is result. And the result in their network with their complexity is 15% better utilization.

Now they can turn that into merit value. The other interesting thing is the 26 hours for only having 30, now they're down to the last trip and I did in Japan, they're telling us they can do 1,000 less than 30 seconds, 26 hours, 30 seconds. Those are the things I look for. And you see Ford Otosan, Pattison Food Group, Shionogi, we talk about it, all example. Now as the revenue sales guy, what I'm looking for is any example of companies making big commitments. And that was what we achieved last quarter, where we had a Fortune 100 companies that after getting results of the first use case, a POT. So wait a second. That's going to go on. But if we can get the results with that, we want to go massive across and looking at all the other use cases that we're looking at. And so we have a $10 million deal for 2 years to take as many use cases as they can in production.

And so that came up with a recipe that we're looking for. We're looking for people that already know exactly this is a hard problem. They're not asking for quantum. Hey, this is the problem. I got the data ready. I know exactly the business case I'm going to get for it. The management team, the technical people, everybody is on board because they understand that. Then we make sure, okay, let's make sure that all the data sets, the full data set is ready to go, make sure the compelling value is there. Do we have a technical folks or business folks all ready to go? And that the adoption doesn't require a major change. If I do that, all those 5, that's how you get to these kind of deal sizes.

And this is one company that hopefully soon, they're going to be talking about it that can achieve these kind of results, and those are the companies I'm looking for to prove out the model. Now interesting enough, these -- all these people are using what we call QCaaS. They just need a fast computing engine. Since the last year, 1.5 years, we have seen a huge demand for people that actually want to buy a system. 6 major categories of people. Research labs. This is like Jülich in Germany. That was the first that kind of triggered this whole wave for us because if you own the system, you can access to a bunch of more controls on it, we call it research controls, so they can do a lot of things that you can't do with a production environment and push the envelope in research.

The same thing with research university. We heard about the FAU business cases. We have more universities that want to get more into this. We also have seen national quantum programs. We announced one in Italy where they wanted to have us come from, obviously. And there's going to be a quantum computer installed very much where I came from. Lake Como is my family is from and it's going to be located there. The same thing you saw an MOU we did with Korea. Again, all the companies -- countries around the world are thinking, how do I establish this expertise in this quantum technology.

And then given results that Trevor covered, we're getting interest from people that are into AI space and blockchain to see if we can get these kind of results, we want a dedicated machine or machines to get some of that. And last but not least, in highly secure environment like defense, for security reasons, they want systems. So we're seeing demand for both Quantum as a service and as well systems. And so far, I'm talking about annealing, stuff that is a little right now and so on. But we are now getting inbound interest. We haven't even announced our offering yet in terms of accessing to our gate model quantum computer. I get a company comes to us almost weekly saying, "Hey, we are spending money with other vendors in gate, but we don't have much to show for." We're really interested in this dual rail approach and be able to develop error-aware applications that are -- we can't do with any other quantum gate model quantum platform.

So in this area, we're seeing folks that want to buy a system. We're seeing people that want access to the QCaaS. And you've seen in Q1 already in the gate model market, we're seeing paid research, which is a different thing that we never had as an annealing beforehand. So that kind of covers my quick overview. Of course, I'm available to answer any questions afterwards. We'll be part of a group. Hopefully, you get a sense. It is probably the funnest job I've ever had because where can you be in a place that you have a technology that customers get excited about it. They're actually willing to come and talk about it. And the light bulb goes on at that POT, when they see their data and their results, all of a sudden, they light up. It's a fantastic moment, and I'm looking forward to having many more.

So thank you very much. And now John can talk about the numbers.

Please welcome to the stage, D-Wave's Chief Financial Officer, John Markovich

Good afternoon, everybody. Thank you so much for joining us this afternoon. We really appreciate the allocation of time and the travel it took to get here. I'm going to highlight on several elements of our revenue and then share some metrics, some of which you've seen before perhaps and some of which you have not seen before.

So let's first talk about our revenue model. So we derive revenue, and this is historically, we derive revenue from 4 particular or specific products or services. As you've heard earlier today, last year, we started to commence the sale of systems. We sold our first system to Jülich Supercomputing Center in Germany last year. And in the first quarter of this year, we booked a $20 million system sales transaction with Florida Atlantic University. And as you just saw from Lorenzo, we've got a lot of interest in that area. So that's a relatively new category of revenue for D-Wave. QCaaS, we have been offering access to our quantum systems through our Leap Cloud system since 2018. This is where the vast majority of the number of our customers are engaged. And then we have the professional services organization that Lorenzo outlined the various steps there. The principal focus of the professional services organization is what we call a means to an end.

Our objective is to utilize professional services as a pipeline to get customers into long-term production QCaaS contracts. Lastly, we have some other revenue, not of the order of magnitude of the first 3. We have service and maintenance on the systems that we sell. These are typically multiyear contractual arrangements. And then we also provide a pretty broad curriculum of training. This is for our customers and for our prospects. And actually, this is all available online. So this is available to basically the general public. We also have a number, and you've heard us talk about this today of professional or potential revenue opportunities going forward. AI, blockchain and then you've heard us talk a little bit about the opportunities within government.

Historically, D-Wave has not recognized much, if any, material revenue from sales to the United States government or any agencies. That is starting to change in a very significant manner, particularly given the highlights of the blessing, if you will, that we got last week from the United States Department of Commerce, along with a capital commitment. And then lastly, the very vast majority of our revenue and our customer engagements have been through our direct sales organization. We are now to the point where we are starting to expand those touch points through channel partners, particularly ISVs and systems integrators. This is a composition of our revenue cut by 3 different ways: by geography, by customer type and by product type. The time frame here is the last 2 fiscal years and the first quarter of this year. As you can see with respect to geography, we've got a pretty substantial footprint in Europe.

Revenue type or by customer type, pretty diversified across research, government and commercial. In our most recent quarter, our commercial customers represented more than 50% of the total revenue. And then revenue by product type, you'll see last year, there's a very large orange portion of the bar here, which is system sale. That is the sale of our first system to the Jülich Supercomputing Center. The reason that bar or a portion of the bar is so large is that our targeted sales price for our annealing quantum computers are in a $20 million to $40 million range. The chart to the far right is what we think that it's probable that the revenue mix by product type is going to look at scale. And we're defining scale as the point at which the company becomes profitable. Notice there's no numbers on that, okay? Because I know I was going to be asked that question.

So we're anticipating based upon the interest that we're seeing on the systems side that revenue mix should be roughly 40% QCaaS, 40% systems, about 15% professional services and the balance being the other revenue categories that I defined earlier. Revenue recognition. So our -- the manner in which we're recognizing revenue varies pretty significantly across our various products and services. The one that's probably the most complex that we get the most questions about is on our systems sales. When we sell a system and we install it, there's a variety of different steps necessary between the actual physical delivery of the system and when it becomes fully operational. The revenue recognition is what's called a percentage of completion.

So through that process, we are recognizing revenue through the process up until the point in time that, that system is turned on, fully tested and is fully operational. That could range from a couple of months to a couple of quarters. Within QCaaS, we now have 2 broad types of QCaaS contracts. One, which I'll characterize as our regular our historical QCaaS contracts are contracts with customers that are accessing our Leap Cloud system for a variety of different reasons. It's a subscription-based contract. So it's very similar, if not identical to a SaaS revenue model. We recognize the revenue on a straight-line basis or ratably over the term of the underlying contract, which can range from a couple of months to a couple of years.

Last quarter, with the enterprise license agreement that Lorenzo just outlined, we have a new type of QCaaS model. We're calling this an enterprise license, and it actually incorporates 2 elements of revenue. It incorporates professional services and it incorporates QCaaS. That particular contract for $10 million is a 2-year contract, and it will be recognized on a straight-line basis over that 2-year time frame, and that commenced in the first quarter of this year. Node placement. This is something we haven't talked a lot about publicly, but I thought it was warrants mentioning because I think we're going to see more of these types of arrangements going forward.

The contract that we announced last year in Italy, the Q-Alliance contract, it's a EUR 10 million contract. This is what we are calling a node transaction. And what we mean by a node transaction is we place a company-owned quantum computer at the customer site, and we do this on the basis that, that customer has contractually committed to buy a significant portion of the compute capacity of that system over a multiyear period of time, typically for a defined field of use. Lastly, professional services, Lorenzo just took you through the 3 preproduction steps in advance of an application actually transitioning into production.

Each one of those steps, which is roughly averages about 3 months, again, we're utilizing the percentage of completion. So if you've got one of those steps that takes 3 months or 4 months, the revenue associated with that step, such as the proof of concept is recognized on a percentage of completion based upon what the contract value is for each step. An example of that is our engagement with Shionogi, and you heard reference earlier to a large major global airline that we're currently working with. So those are both the examples of professional services type of agreements.

Key metrics at scale. Remember how I define scale. So these are the gross margins that we are anticipating. QCaaS, 65% to 75% professional services in the range of 40% to 50% and then quantum computing systems between 75% and 80% or 90%. And the variability there could have to do with the actual specifics of that type of contract because there could be different types of elements associated with that. With respect to operating expenses, we are going to continue to invest very heavily in research and development to support both the annealing program as well as the gate model program as well as our investments in the areas of AI and blockchain. Sales and marketing, we will continue to expand our capabilities on a global basis. And the net of that is the general and administrative expenses.

Given that we now have 2 substantial research and development centers, the first one being in Burnaby, the most recent one being in New Haven, Connecticut through the acquisition of Quantum Circuits and the one that's on the way in Florida, which will be located in Boca, there's going to be an amount of capital expenditure that's going to be necessary on an ongoing basis to fully facilitate those 3 locations. So we're anticipating that our annual recurring CapEx investment is going to be in the $15 million to $25 million range. We're frequently asked what our capacity is, our QCaaS capacity. Each one of our production systems can support between $25 million and $30 million of annual QCaaS revenue.

As Trevor highlighted earlier, we have 4 production systems that are supporting the LEAP cloud system today. So that translates to about $100 million to $120 million of annual revenue capacity. So there's a tremendous amount of operating leverage inherent in that. To the extent that we want to expand that capacity, it costs us approximately $2 million to build, calibrate and install an annealing quantum computer over roughly a 4-month time frame. From a liquidity perspective, we ended the first quarter with $588 million in cash. During Q1, we invested $250 million of our cash in the acquisition of Quantum Circuits, along with the issuance of common shares as well.

Over the last 9 quarters, we have raised slightly more than $1 billion in equity. Roughly 70% of that was derived from curious when we reach different ATM programs. Last year, we had nearly -- by the end of the year, 100% of our warrants were exercised. That contributed about $203 million to the balance sheet. And the balance is made up of closing out or topping off the equity line of credit program that we put in place when we first went public several years ago. And then there was contributions from the employees exercising their options.

You've heard some references today to our patent portfolio. This quantifies the magnitude of the portfolio. D-Wave has consistently been ranked as having one of the largest quantum computing patent portfolios in the industry. Through the acquisition of Quantum Circuits, we just added in a very significant manner to that. So we have over 270 U.S. granted patents. We have about another 320 that are pending in the United States and internationally. So that provides us with more than 590 issued and pending patents. And through the acquisition of Quantum Circuits, we became the exclusive licensee to a very substantial patent portfolio at Yale University, much of which Rob was very involved in patenting, and that is a patent portfolio in excess of 220 patents. So we now have total patent count of over 800.

We categorize our patents into 2 principal categories by type. And as you can see here, 70% or 82% of our patents are systems-based patents and the balance are software patents. And then we also characterize or categorize our patents by the quantum computing architecture. So 46% of our patents support both the annealing architecture and the gate model architecture. 31% are exclusive to the gate model architecture and 22% are exclusive to annealing.

We are fortunate to have 14 security analysts that follow the company and regularly publish on them. I'm honored that most of them are sitting here in front of me. 8 of the 14 analysts are ranked in the top 1% of over 12,000 analysts that are ranked by tip ranks. And as you can see here, I have taken the consensus numbers from all 14 for the price target, the estimated revenue for fiscal '26, which is a little under $43 million and the estimated EBITDA loss, which is approximately $118 million.

Lastly, there are a lot of bullet points on this slide in terms of investment considerations, but I want to highlight a handful that are entirely unique to D-Wave. One, as you heard earlier today, we are the only dual platform quantum computing company, which means we are the only quantum computing company that is positioned to address the entire target market, the TAM. Second, we have developed and we operate the world's most powerful and largest quantum computer. That's the advantage to annealing quantum computer. Third, we have demonstrated quantum supremacy on that system, and that has not been successfully challenged. We demonstrated that on a real-world problem, not a contrive problem, which constitutes most of the other claims to quantum supremacy. We are the only quantum computing company that has got applications in production, meaning that our customers have actually incorporated access to our quantum computers in their day-to-day workflows.

And lastly, 100% of our revenue is derived from either the sale of quantum computers, access to our quantum computers or services associated with our quantum computers, not other revenue categories that might have the word quantum attached to them, such as quantum sensing or quantum networking. Thank you.

Okay. Now John and I are going to tag team for the last Q&A session. I'll let you select, John.

Troy?

It's Troy from Cantor Fitzgerald. Maybe a question for John, if you can share what are the technical milestones that you guys have to reach to capture all of the $100 million investment from the U.S. government?

I'm actually going to ask Trevor to address that question because he was very intimately involved in developing those milestones.

So it's a mix of tooling that we want to actually buy for our own research foundry facility as well as tooling that we will put into our production foundry partners. There is fabrication development and part of the milestones are delivery of prototypes that basically showcase the results of that fabrication development. So there's in all about 9 initiatives, including a tooling initiative and milestones over the 5 years that unlock the funding for the CHIPS office from the CHIPS program.

Now broadly, it's in the areas of things that we are pushing on and developing right now, including the high-density interconnects that Alan spoke about, the superconducting interconnects to connect integrated circuits together, novel high-performance dielectrics for our annealing fabric to drive up coherence times, new wiring and interface engineering to really drive up coherent and multilayer stacks and then some fundamentally new fab techniques for driving potentially higher performance and Jülich technology. So the details of all the milestones haven't been made public yet, but there's a series of basically tool installs, fab process nodes and prototypes that get delivered that make use of those process nodes.

Well, now they've been made public.

Tyler Anderson from Craig-Hallum. I just want to add on to that. And my main question is, for the tooling, is this 300-millimeter capacity or at least additions or add-ons to tooling that you can purchase?

So our current tooling is in our current node is 200-millimeter tooling. And so we don't have any immediate plans to move away from that. So it's really focused on how do we enhance the process at that wafer scale.

Daniel Stevens with RF Lafferty. My question is more about when you go through your milestones and your chips become more advanced, you built some of your computers on the campuses of other clients. How easy is it to replace those -- the previous chips with your new ones? And how are those clients affected with spending upgrading those computers?

Do you want me to take that? So it actually depends on the generation. Some generation transitions just require a chip change. Others require chip change and maybe some control and I/O changes. But they all use the same refrigerator. At least they all have used the same refrigerator up until now. So for the most part, it's actually pretty straightforward for us to do an upgrade. We warm the system. It maybe takes a week or 2 weeks to install a new chip, any new control or I/O and then cool it back down and calibrate it. So it's actually pretty straightforward.

Gentleman in the white shirt.

Vijay from Mizuho. Just a quick question. On the same $100 million government funding, is that something that you can achieve on Advantage2? Or do you need to get to Advantage3 or multichip Advantage3 to get the entire $100 million? And then back on the POST-NISQ roadmap that you showed in the FAULT-TOLERANT roadmap, the POST-NISQ showed the physical qubits, but not the logical qubits. And on the FAULT-TOLERANT side, I thought you guys showed the logical qubits, but not the physical qubits that you need to get to that. So if you could just hit those two.

I'll take the second one first and then Trevor can answer the first question. Yes, the reason why we didn't really talk about logical qubits in the POST-NISQ arena is because, a, there's not really enough error correction to be able to consider them error-corrected qubits and b, the number -- the total number of physical qubits is small enough that really what we're talking about is like one logical qubit on the chip, right?

So it's more about the error detection capability and new algorithms using the error detection capability and then our ability through our own scientific research to show what the fidelity of that one logical qubit would be. It's not do we get into the FAULT-TOLERANT with 1,000 and 10,000 qubit systems that we have multiple logical qubits.

So I can address the first part of your question, which is the kind of, again, getting at the deliverables for the CHIPS program. So there's multiple initiatives that were funded under the CHIPS program. One of them, for example, I just spoke about is improved dielectrics. So dielectric materials that have much lower loss, so that should drive higher coherence. To the extent that those dielectrics become available as we develop them, we will incorporate them into our products, including Advantage2 over the next couple of years.

But another milestone, another initiative really is, again, around the multichip packaging. So -- and that, obviously, the prototypes would need to be multichip prototypes that exercise that packaging technology.

I think it was communicated, but I'm not sure. It is a 5-year program from start to finish. And the funds are released as the various milestones are achieved.

Kingsley Crane, Canaccord. If we think about gate model, as complex as the technology is it's still -- the go-to-market is a little bit more one dimensional. It's sort of like build it and they will come. For annealing, it's a bit more dynamic. I mean you have 4 systems out there. The ceiling is $120 million spend, the ARR is still a little bit lower today. So on the technology, you can do a lot with 2 million variables. We've seen that with the likes of AT&T. You can do a lot more with 50 million. So I'm curious when we reach 50 million? Is that 2029 or 2031 or later?

And then in terms of the go-to-market, you have that framework of different gating factors within Leap. And we've talked about data availability is a big roadblock. But what can we really do to unlock that, whether that's data availability or other parts within Leap to kind of unlock and get closer to $120 million?

Yes. So on the annealing side, it's an ongoing process, right? I mean, part of the reason why we have a limitation on the number of variables right now with the current solvers and the current processor is because remember, what the hybrid solver is doing is it's trying to essentially find hard sub problems and send them off to the quantum computer. And right now, we're limited on the size of the sub problem, and that limits the size of the total problem. So as the quantum computers become larger, the size of the sub problems become larger, and that allows us to solve even larger total problems. And that's an ongoing process.

At 100,000 annealing qubits, would we be able to support a 50 million variable problem? I would say probably not. We still need to keep growing, but two things will happen at 100,000 qubits. One, we will be growing the number of variables that we can support. And two, for the problems that we even could solve on the 4500 system, we'll be getting even better solutions because we're able to let the quantum computer do even more of the heavy lifting.

What was the second question -- the data pipeline and what can we do to help address that. So part of what we are trying to do is to have those discussions early on with our customers so that we identify problems right upfront that are not only challenging for them, but for which they have an environment that if we are successful, we can sooner rather than later move into production or if they don't, at least the right people are engaged to start thinking through and working through how to do that.

So in the early days, like now, it's really all about making sure that we increase our probability of success in getting to production so that we can get a reference-able customers that can then help us communicate to other customers the value and how to go about evolving your infrastructure to support this.

I think another question here. For you, Alan. If you look at the road map, it looks like you guys are assuming about 100:1 physical to logical ratio. I guess I would have thought given you guy's better fidelity that you guys would have a much better ratio like that. We know we've got one company out there claiming a 2:1 ratio. So can you just talk about how good we can get to and what type of ratio can get to?

2:1 ratio. Okay. Don't get me started. Rob, do you want to talk about 2:1 ratio or Robert.

So we're being conservative there, I would say, in the ratio of physical to logical qubits. We're kind of showing you projections for the best known code, the surface code, which is not necessarily the most efficient. But you have to be careful a little bit about some of these newer codes where people know that if you can encode, it's more efficient, but not yet like how you do operations or -- so if they don't have a complete set of operations defined for that, it's not so clear to me whether you can really realize that amount of low overhead.

I guess the way I would say it is that there are approaches to error correction that are well understood and for which you can have -- you can kind of develop a viable road map. And then there are approaches that are great on paper, but not all that well understood and for which there are huge challenges that remain if we can ever overcome the challenges. So when somebody says, I'm going to be able to error correct to 6 or 7, 9s with 2 physical qubits per logical qubit, maybe in 20 years, I don't know.

I think we'll go online. Yes, there's a couple actually. The first is how long does it take to build the system from start to revenue production? And then do we actually put systems in inventory? Do we build them to order?

Yes. So if we have all the parts, then it generally takes on the order of 2 to 3-weeks to actually build the entire system, then we have to cool it down, and it can take a week to get it cooled down, and then we have to calibrate the processor. And the calibration is what takes the bulk of the time. Calibration can take 3 to 4 months. At that point, the system is operational.

So we basically say 3 to 6 months from start to finish to have a fully operational system. As far as the parts are concerned, we're constantly evaluating our supply chain. We do a quarterly report on the supply chain. Today, we have no problem with access to all the components that we need to build the systems. And in the past, when we're primarily a quantum compute as a service business, and so we don't didn't need a lot of systems, we weren't ordering in advance. Now that we're moving into a more system sales model, and we've got what looks to be like a very, very strong pipeline, we are now preordering to be able to build as we sell.

The second question is regarding the customer examples Lorenzo talked about, how much custom engineering is required? And is any of that transferable to other customers in the same industry?

I don't know, Lorenzo, do you want to answer that question?

Yes. Well, the formulation is very specific to the customer data set. Of course, you get expertise. And if you look at manufacturing problems, you start seeing repetition across. Sometimes we come up with an area that Trevor's team jumps in, and that's an extension. And those goes into the core products of the next customer has the same benefit.

For example, the most recent one, if you want to improve a variable that is driven by a machine learning algorithm, for example, that we can embed it in, and now that's available for every customer. So that's kind of the general path. And so in general, we're seeing now we have a team focused on -- you get expertise in one class of use case and figure out, okay, what are the customer, what other industries of similar ones. For example, scheduling, service repairment that go and have to show up with parts and so on. guess what? There is a lot of similarities there in, let's say, the airline industry. So any kind of -- so you start seeing those similarities. In general, we're organized by industry, so we can kind of leverage those learning curves.

Craig Ellis at B. Riley Securities. So I wanted to ask a finance-related question. So thanks for all the inputs to the scale model. And I would love to ask what's the revenue number, but I'll ask the question in a different way. And I'll put it this way. I'm sure you're looking at a number of prerequisites across the different businesses to get there, whether it's utilization levels with your 4 or more QCaaS systems or certain points on Rob's road map or number of system sales. Can you just help us understand how you as an executive team and in your discussions with the Board are looking at those issues so we can have a better sense of the path from here to there?

So is your question an actual revenue question? Or is the path how to get there?

Well, do you want to give me the revenue...

I'm sure you would. I thought I addressed that earlier. Craig, is your question capacity to get there? Is your question capacity to get there?

Well, utilization on the [indiscernible].

I heard it, John, more as, okay, if at profitability, it's 40% systems, 40% QCaaS and then some of the other stuff. How many systems is it going to take to get us there, which, of course, if we tell you, we'll also mathematically get you all the data you're looking for.

And if I gave you the capacity utilization on the QCaaS, which I've already given you that number, you can kind of back in.

So look, here's what I'll say. And this is kind of a repeat of what I said on the earnings call. When we first started talking about System sales, I said, think 1, a year, right? A year later, I said, think 2 to 3, a year. But I'm seeing a really strong pipeline right now, and I see that growing nicely into the future. And so I think if you start thinking about like natural progressions off of 1 to 2 to 3 to -- you can probably figure out how to get there.

So this is for Rob and Trevor. Non-Clifford gates are what give quantum computers all their power. So I'm really curious about the overhead associated with magic state factories in your road map and how you address that overhead?

So full marks, yes, this is correct. So there's been a lot of recent theoretical progress on how to reduce the overhead associated with [indiscernible] state production and the like. And one of the things that's interesting and pretty exciting there as well is the idea that you -- of course, when you're preparing these states, you get to cheat a little bit. You can use repeat until success and detection. So something that we're looking at more closely is like how far up the ladder of fidelity we can be as a starting point for producing those states. So I think that's another place where you can see a good bit more progress potentially coming from that.

So it's almost on the software side [indiscernible]

I think it's a combination of both. So it's the point that we have the hardware that can place some new tricks and the learnings that people are making on the theoretical side.

Maybe just one more question. Part of the U.S. government's recent investments included $1.35 billion to secure the world's first Quantum foundry with IBM and $350 million at Global Foundries. You've had one of your foundry partners in the process of being acquired. Can you just share your thoughts on would you be willing to use IBM as a potential customer -- sorry, competitor? Would you be willing to use Global Foundries? Just any thoughts on the process technologies of those two companies?

I mean we didn't know ahead of the announcement by the U.S. government who would be getting funding in this round. But as soon as I heard the announcement, I send an e-mail to Trevor and I said, can we use the IBM Foundry? Absolutely. If they have technology that we can make use of, and it's open to other potential companies to leverage, which I suspect it will be given that it's got a significant U.S. government investment, then we would absolutely use them. I think that there are some owners of foundries that I have a bit more comfort working with than others. And yes, I mean, if IBM has got technology that we can leverage and the U.S. government is a significant investor and partner, I would be very open to using it.

Okay. I think that, that wraps for the day. We do have some refreshments out in the entry way, and thank you all for joining us. You have a good day.

Good morning, and welcome to the D-Wave First Quarter 2026 Earnings Call. [Operator Instructions] Please note this event is being recorded. I would now like to turn the conference over to Kevin Hunt, Investor Relations. Please go ahead.

Thank you, and good morning. With me today are Dr. Alan Baratz, our Chief Executive Officer; and John Markovich, our Chief Financial Officer. Before we begin, I would like to remind everyone that this call will contain forward-looking statements, which are subject to risks and uncertainties and should be considered in conjunction with cautionary statements contained in our earnings release and the company's most recent periodic SEC reports. Both an on-demand webcast and a transcript of the conference call will be available on the Investor Relations section of the website within 48 hours after the call.

During today's call, management will provide certain information that will constitute non-GAAP financial measures under SEC rules, such as adjusted EBITDA loss and non-GAAP adjusted operating expenses and operating metrics such as bookings. Reconciliations to GAAP financial measures and certain additional information are also included in today's earnings release, which is available in the Investor Relations section of our company website at ir.dwavequantum.com. This morning will be limited to taking one question from each analyst during the first round of questions and then time permitting proceed to a second round of questions where again, we will have to limit analyst I'll now hand over the call to Alan.

Good morning, everyone, and thank you for joining us. I want to begin by sharing a few perspectives on the quantum computing landscape. As Lynas Turbus, the creator of Linux, once said, Talk is cheap, Show me the code. That idea feels especially relevant in quantum computing today as excitement rises, competition increases and investors are looking for proof, not just promise. As quantum computing attracts increasing investor attention, the sector is generating significant excitement but also significant noise.

New entrants, new public listings and ambitious and some might argue very overstated technology and product claims are drawing attention. In this environment, it is more important than ever to distinguish hype from execution. As the field grows more crowded, the conversation is shifting from who is participating to who is positioned to deliver. Customers are looking for technology that can give them a competitive edge today. Researchers want cutting-edge tools that can accelerate discoveries now, and investors are working diligently to separate signal from noise in a sector that can sometimes generate more headlines than evidence or results.

As the CEO of one of the world's first and leading quantum computing companies, I believe that it is important to help educate the market on the realities of quantum computing's true near-term capabilities and commercial traction as well as the strengths and trade-offs of the different quantum computing modalities as this market continues to take shape.

Let me be direct. Many still view e-Wave through an outdated lens, but I think it's time for a vision check. We believe we are the clear leader in quantum computing today, a market that Boston Consulting Group, BCG, projects to be in excess of $800 billion. A full 1/4 of that market is optimization, which is uniquely addressed by annealing quantum computing, where we are the only player. This is not a niche technology or market as some like to characterize it. It is estimated by BCG to be a whopping $100 billion to $220 billion opportunity. For comparison, that's as big as either the global semiconductor equipment market or the global cybersecurity market. For those who say that D-Wave is addressing a niche market, stop spreading competitive misinformation and start doing your homework. Do you think that the global semiconductor equipment and cybersecurity markets are niche? I don't.

What's more, D-Wave is now also a leading player in gate model quantum computing through our acquisition of Quantum Circuits in January of this year. This means that we are the only company in the world that can participate in the full addressable market for quantum computing with both annealing and gate model quantum computers. We are building a highly differentiated pure-play quantum computing company with proven commercial traction today. For those valuing us on only our annealing technology and products, I would say you're off by a factor of two. Our unique market position reflects our rapidly advancing gate model progress, which has greatly accelerated given Quantum Circuit's industry-first dual rail qubit technology.

It is imperative that you all understand the profound potential that this has on our business and ensure that the models you're using to evaluate the company take into account this newly acquired technology. With dual rail qubits built-in error detection, we believe that our gate model quantum computing systems will set a new standard on quantum performance, efficiency and scalability. Our dual rail gate model technology is a meaningful differentiator for D-Wave and in our view, one of the most important developments in quantum computing today. It brings together super connecting speed, the fidelity associated with ion traps and neutral atoms and a clear path to scale with our proprietary on-chip cryogenic control technology, a combination that the market should be paying much closer attention to.

This combination is a revolutionary approach to the development of a gate model quantum computer. Trucked ions or neutral atoms are like a bicycle. They're simple, reliable and efficient, but very slow. Superconducting is like a piston propeller airplane, much more complicated and far less reliable, but much faster. In fact, there was recent research from Google and then separately from John Presco at Caltech and Atonic that brings this to life.

The Google team showed a way to break the Bitcoin protocol with 500,000 superconducting qubits. Then the Presco team showed how you could do it with only 10,000 neutral atom qubits, much more efficient. But what wasn't highlighted was that the Google computation would take about 9 minutes on a superconducting quantum computer. This computation would take many months, months on a neutral atom quantum computer. At that speed, they really aren't breaking the Bitcoin protocol. But this demonstrates the point that superconducting is much faster, actually 1,000x faster, but less efficient, requiring more physical qubits to error correct.

Well, D-Wave dual rail qubits provide the best of both worlds. Again, the best of both worlds. Think a jet airplane. -- still much faster than trapped ions or neutral atoms, but much more reliable and efficient than an old piston airplane. You get the speed of superconducting with the efficiency of ions or atoms. This is truly revolutionary. And the implications are significant. more reliable computation, more efficient error correction and a potentially faster, lower overhead path to building useful quantum systems. With built-in ratio detection, these qubits can identify roughly 90% of errors as they occur with an observed erasure rate of just 0.5%. We have also demonstrated greater than 99.9% fidelity while reducing the number of physical qubits needed for a logical qubit by up to an order of magnitude.

Our advantage becomes even stronger when dual rail is combined with Vie way's proprietary on-chip cryogenic control. This gives us a path to significantly reduce the wiring required to control large numbers of qubits and ultimately enables full qubit control at scale with multiple orders of magnitude fewer control lines than competing superconducting gate model systems. Any technology that doesn't solve this issue will not achieve utility because it can't feasibly control without requiring football field-sized installations. The combination of speed, fidelity and scalability is what makes this such an important development. It is not just better qubit design. It is a more scalable system architecture, which is why you should see dual rail technology as a clear competitive advantage for D-Wave.

I'm excited to share with you today more visibility into our gate model road map. We are targeting our dual rail gate model road map to extend to 100 logical qubits by the end of 2032. By the end of 2028, we plan to have approximately 175 physical qubits, which will allow us to demonstrate our quantum error correction technology as well as logical operations. Beyond this, the integration with D-Wave's scalable control is expected to take us to 10 logical qubits by 2030, followed by 100 logical qubits 2 years later. This acceleration in our road map is based upon the unique opportunity provided by the recent merging of Quantum circuit expertise in engineering high coherence superconducting quantum devices with D-Wave's extensive toolbox for scaling superconducting quantum processors. With 100 logical qubits, we expect e-Wave to capture as much of the gate model market as any other gate model quantum computing company.

There's something else that investors need to see more clearly about road maps in this industry. You've seen companies revise time lines, change milestone frameworks and move the goalpost as the technical realities and complexities of scaling become clearer. We couldn't be more different. The road map we are sharing is built on demonstrated technology, known engineering pathways and milestones that we believe are achievable with a high degree of confidence. We are not publishing dates for effect. We'll provide more detail on our product road map and how it compares to other gate model quantum computing modalities like neutral atoms, tracks and photonics at our upcoming Investor Day on June 1 at the NYSE and online, and we encourage you to attend.

Our category leadership position is further solidified by our dominance in annealing quantum computing, clearly a foundational strength for the company. This is grounded in a long track record of innovation and product delivery across 6 generations of systems. Our annealing quantum computers are being deployed today in real customer problems. They are trusted by some of the world's largest companies across manufacturing, aerospace, health care, telecommunications and other sectors as well as by leading scientific researchers using our systems to accelerate discovery. This is real work, driving real value right now.

Beyond optimization, we're very excited by what we're seeing in the area of blockchain. We recently collaborated with Postquant Labs on the development and launch of its quantum classical blockchain testNet, which is now live. The testNet is designed to support the development of a global quantum blockchain standard and to assess how quantum computing could contribute to a more secure and energy-efficient blockchain in a distributed network. More than 18,500 people have signed up to participate in the TestNet. It currently includes more than 1,600 nodes, one of which is D-Wave's Advantage2 annealing quantum computer with the rest made up of CPUs and GPUs. Our Advantage I QPU is currently outperforming the classical nodes and winning the majority of the blocks.

Together with Postpot Labs, we are launching a detailed benchmarking study to further quantify the advantage. We're also seeing promising work in the area of quantum AI and machine learning. Shionogi, a Japan-based pharmaceutical company, is working on a multistage progress project that applies AI to drug discovery, where identifying drug-like molecules with the right activity, chemical properties and synthetic accessibility is extremely challenging, particularly for classical machine learning methods. The work is focused on using D-Wave's annealing quantum computers as part of the large language model training process with the second phase of the project delivering a tenfold increase in the number of desirable molecules compared with the results generated using a classical machine learning algorithm.

Shionogi is now moving into the next phase of the project with the ultimate goal of real-world adoption. We believe that these early results, along with emerging work by other customers exploring quantum computing to improve AI performance, position Z-Wave as an important first mover at the intersection of quantum and AI. Together, these examples show that annealing quantum computing is expanding commercially, opening new application areas and continuing to demonstrate real-world value today. Not only can our annealing quantum computers uniquely address the significant and important optimization market, we are close to being able to demonstrate their value in AI and blockchain.

We continue to expand the capabilities of our annealing quantum computers. We recently published research outlining powerful new multicolor annealing protocols that enable some gate model operations within our commercial Advantage 2 systems. We also launched these features with key customers to enable them to perform fundamental research in quantum simulation. These protocols enable researchers to use D-Wave's annealing GPU to model quantum systems and explore fundamentally new behavior that can be extremely difficult, if not impossible, to study with classical techniques.

On our last quarterly earnings call, I said that 2025 was an inflection point for D-Wave, and our results continue to support that view. Last year marked a period of clear technical progress and accelerating commercial momentum, including a triple-digit increase in our sales pipeline that continued to expand through the first quarter. Today, that momentum is translating into measurable business outcomes. As discussed, in January alone, we signed 2 landmark agreements, a $20 million system sale to Florida Atlantic University and the industry's first $10 million enterprise license quantum computing as a Service deal. We have previously covered those transactions, so I won't repeat the details here, but I will emphasize their impact as they help to drive record first quarter bookings.

During the first quarter, we closed bookings of $33.4 million, a nearly 2,000% increase over Q1 bookings a year ago and up 149% from the very strong bookings in the fourth quarter of 2025. With regard to system sales, I also want to point out that while I originally shared with you that we expect to sell 1 system per year, the pipeline is much stronger. We're now expecting 2 or 3 system deals per year with expected delivery of at least 2 systems this year in 2026.

Before I hand the call over to John to provide deeper details on our first quarter results, there are 5 key points that I want you to keep in mind about what makes V-Wave different. First, D-Wave is the only dual platform quantum computing company. We are developing both annealing and gate model quantum computing systems, which we believe uniquely positions us to participate in the full addressable quantum computing market over time. Second, Anealing quantum computing is better suited for optimization than gate model quantum computing. By its nature, it is uniquely built to solve optimization problems, an area that represents a significant share of the overall quantum computing opportunity and one where D-Wave is exceptionally well positioned to lead.

Third, our customers are using our annealing quantum computing systems in production right now. They're solving hard computational problems that directly affect operations. This is not experimentation. It is commercial deployment by several Forbes Global 2000 companies. Fourth, Z-Wave is the first company to solve a hard computational problem beyond classical computing's reach on a real-world useful problem, evidenced by our quantum supremacy results published in Science last March.

And fifth, our dual rail date model technology changes the game. It combines super connecting speed, high-performance fidelity and a clear path to scale in a way we believe is highly differentiated. It is increasingly clear that the winners in quantum will be the companies that combine technical differentiation, commercial proof and the ability to execute at scale. We believe D-Wave is one of those companies, and our first quarter results, along with our momentum in the second quarter reinforce that position. With that, I'll turn it over to John.

Thank you, Alan, and thank you to everyone taking the time to participate in today's call. Revenue in the first quarter of 2026 was $2.9 million, a decrease of $12.1 million or 81% from the first quarter of 2025 revenue of $15 million that included $12.6 million in revenue from the first sale of the D-Wave annealing quantum computer system. For the first quarter of 2026, D-Wave recognized revenue from over 100 individual customers, over 50% of which were commercial enterprises with commercial revenue constituting over 73% of the $2.9 million in quarterly revenue.

From a product perspective, Q1 revenue was comprised of $1.8 million in QCaaS subscription revenue that increased by nearly 15% on a year-over-year basis and $1 million in professional services revenue that increased by over 26% on a year-over-year basis. For the first quarter, 100% of Gateway's revenue was derived from selling, providing access to or providing services for quantum computing systems, not other revenue that has the word quantum attached to it, such as quantum sensing or quantum networking.

Bookings for the first quarter were $33.4 million, an increase of $31.8 million or 1,994% when compared to the 2025 first quarter bookings of $1.6 million. and an increase of $20 million or 149% when compared with the immediately preceding 2025 fourth quarter bookings of $13.4 million. Over two dozen commercial customers comprised over 31% of the first quarter bookings with the balance of the bookings from educational and research organizations, the largest of which was the $20 million system sale to Florida Atlantic University.

During the first quarter of 2026, the dollar value of our sales opportunity pipeline more than doubled over the dollar value of the sales opportunity pipeline as of the end of the immediately preceding fourth quarter of 2025, while the average potential deal size more than doubled over the same period. GAAP gross profit for the first quarter was $1.8 million, a decrease of $12.1 million or 87% from the 2025 first quarter GAAP gross profit of $13.9 million that was heavily influenced by the aforementioned system sale in the first quarter of last year. GAAP gross margin for the first quarter was 63.6%, a decrease of 29% from the 2025 first quarter GAAP gross margin of 92.5% that again, was heavily influenced by the aforementioned system sale in the first quarter of last year.

First quarter GAAP operating expenses totaled $56.5 million, an increase of $31.3 million or 125% from GAAP operating expenses of $25.2 million for the 2025 first quarter. with the increase primarily driven by $9.1 million of nonrecurring costs related to the acquisition of Quantum Circuits, an increase of $8.6 million in salaries and related personnel costs, 80% of which relates to increases in sales and marketing and research and development personnel, including Quantum Circuits operating expenses subsequent to the closing of the transaction in January and $7.4 million in noncash expenses, including $4 million in stock-based comp and $3.4 million in depreciation and amortization expenses. These increased operating expenses stem from investments to support the company's accelerated product development and go-to-market initiatives as well as Quantum Circuits.

Non-GAAP adjusted operating expenses were $34.8 million or $21.7 million lower than the GAAP operating expenses with the non-GAAP adjusted operating expenses increasing by $14.6 million or 73% over the year earlier non-GAAP adjusted operating expenses of $20.2 million, with the difference between GAAP and non-GAAP operating expenses primarily being noncash stock-based comp, noncash depreciation and amortization expenses, in nonrecurring onetime expenses, such as the $9.1 million in nonrecurring costs associated with the Quantum Circuits acquisition that are excluded from the non-GAAP adjusted operating expenses.

Net loss for the first quarter was $18.4 million or $0.05 per share compared with a net loss of $5.4 million or $0.02 per share in the first quarter of 2025, with the increase due to higher operating expenses primarily associated with our increased investment in our R&D and sales and marketing organizations and lower gross profit given the high gross profit associated with last year's sale of an annealing quantum computer. This was partially offset by the increase in income tax benefit of $28.5 million that was derived from the January 20 acquisition of Quantum Circuits. Adjusted EBITDA loss for the first quarter was $32.8 million, an increase of $26.7 million from the 2025 first quarter adjusted EBITDA loss of $6.1 million, with the increase due primarily to higher operating expenses and lower gross profit.

Now I will address the balance sheet and liquidity. As of March 31, D-Way's consolidated cash balance and marketable investment securities totaled $588.4 million, an increase of $284.1 million or 93% from the 2025 first quarter consolidated cash balance of $304.3 million. During the first quarter, we invested $250 million in cash in conjunction with the acquisition of Quantum Circuits, and we believe that our remaining liquidity is sufficient to support a fully funded plan to profitability.

Subsequent to the 2025 fourth quarter earnings call that was held on February 26, I've received a number of questions on revenue recognition that I touched on during our fourth quarter earnings call that I will reiterate here, specifically as it relates to system sales. These transactions involve a number of steps before the systems are fully operational, including site preparation, delivery, installation, calibration and other key steps that are likely to encompass multiple months and possibly quarters depending upon the unique elements of a particular system transaction. While we will recognize a significant portion of revenue upon the physical delivery of the system, we will recognize a smaller portion over time prior to delivery as installation and calibration progress since these activities are essential for customers to begin using our systems.

This is the general pattern we expect that each system sale may have unique characteristics that may cause the revenue recognition pattern to vary somewhat. In addition, we anticipate that most system sales transactions will involve 1 or 2 multiyear revenue components, including a service and maintenance contract and access to our cloud service. In conjunction with touching on the topic of revenue recognition, we thought it would be helpful to highlight the recent progression of our remaining performance obligations or some we refer to this metric as RPOs or backlog.

As of March 31, the aggregate amount of remaining performance obligations that were unsatisfied or partially unsatisfied related to customer contracts totaled $42.4 million. That represents a $36 million or 563% increase over the first quarter of 2025 RPO balance of $6.4 million and a $29 million or 216% increase over the immediately fourth quarter 2025 RPO balance of $13.4 million. Approximately 54% of the $42.4 million first quarter RPO balance is expected to be recognized as revenue in the next 12 months and 71% is expected to be recognized as revenue in the next 2 years with the remainder to be recognized as revenue thereafter. Revenue allocated to remaining performance obligations represents the transaction price of noncancelable orders for which service has not been performed, which includes deferred revenue and the amounts that will be invoiced and recognized as revenue in future periods from open contracts and excludes unexercised renewals. The same information is also included in our Form 10-Q.

While we are continuing our practice of not providing specific forward financial guidance, given the revenue recognition associated with systems transactions, in combination with the remaining performance obligations and the sales pipeline, I want to provide some directional parameters on revenue over the balance of this year. The 2026 second quarter is likely to be up modestly from the first quarter with a substantial portion of the year's revenue recognized in the second half of the year. In conclusion, as we have previously stated, we continue to believe that D-Wave has the opportunity to be the first independent publicly held quantum computing company to achieve sustained profitability and to achieve this milestone with substantially less funding than required by any other independent publicly held quantum computing company. With that, operator, please open the call for questions.

[Operator Instructions] The first question comes from Quinn Bolton with Needham.

This is Shaw on for Quinn. Congrats on the increased system sales outlook. I guess staying on that topic, what's driving your confidence in being able to secure 2 to 3 system sales a year? And how do you view the split between annealing and gate model going forward?

Sure. So as both John and I indicated, our pipeline has significantly increased over the course of the last year. And we are well down the path of negotiating system deals with multiple customers, none of which has been communicated to date. So with the Florida Atlantic University sale this year and the progress that we're making on several other system deals, I have a very high degree of confidence that we'll see 2 or 3 sales this year. And as I said, a very high degree of confidence that we will actually deliver 2 of them this year.

The next question comes from John McPeake with Rosenblatt Securities.

Al, John and Kevin, congrats on the bookings and RPO number, pretty impressive. So a question on the road map here. By the end of 2032, we have 100 logical qubits. Could you give us a sense as to what you're targeting for 2 qubit gate fidelities out there? And I have the same question about the 10 logical qubits in 2030.

Yes. So first of all, we're already at 99.9% fidelity, but that's on a very small system admittedly. One of the things that we believe the dual rail qubits are going to do is put us on a much steeper path to improving fidelities. In particular, the Google Willow work was quite impressive, but we believe that with our dual rail technology, we'll be able to improve upon that by about 5x. And so we're looking at very high qubit and 2 qubit gate fidelities.

The next question comes from Antoine Legault with Wedbush Securities.

Congrats on the momentum so far this year. You've effectively been the sole player in quantum annealing for over a decade. As the addressable market for optimization grows, and Alan, you've cited some pretty significant figures in terms of addressable market as annealing commercial viability becomes more established, do you expect to see more entrants, whether it's from other established gate-based players pivoting towards hybrid approaches or others moving into the space? Like how should we think about the competitive landscape going forward?

Yes. So first of all, I do want to point out that the numbers I quoted for optimization are the Boston Consulting Group numbers. So this is the data that most people in the quantum industry are using and focused on with respect to the total addressable market and the $100 million to $220 million -- billion number comes from Boston Consulting Group for optimization. Second, actually, we're already starting to see others working on annealing systems, very small at this point in time, 2, 3, 4 qubit systems. We're also seeing some gate model companies starting to look at running annealing type protocols within their gate model systems. Sometimes you'll hear a gate model company say they've done some analog computing within their gate model system. Part of the reason for looking at this is that as we've talked about in the past, annealing is far less sensitive to errors and doesn't require error correction to give good results.

But the problem with that is there's a lot of overhead associated with trying to run annealing protocols within gate model systems, and they'll never be as fast or never be able to solve problems as large as what you can solve on a native annealing quantum computer. So yes, there is increasing interest in the annealing approach to quantum computing. There are some early activities underway with respect to building annealing quantum computers, and there is some work going on with respect to trying to perform annealing within gate model systems. But none of that represents a real threat to the advantage that we have in annealing quantum computing. We continue to believe that D-Wave is and will always be the leader in that portion of the market.

The next question comes from Joe McCormick with Evercore.

Congrats on the quarter. Maybe just as you've seen the pipeline progress and kind of expectations for the aperture to start widening as it relates to system deliveries, moving forward here. Can you kind of double-click on that and talk through kind of the appetite for on-prem systems for kind of governments and kind of academic research versus commercial, maybe to the extent that you found there's kind of greater openness on the commercial side of things as well over the next couple of years to taking kind of annealing systems on site?

Yes. So if I were going to guess at what we're likely going to see this year. I think when I talk about 2 or 3 system sales this year, I think that we're likely to see one in the commercial arena and the other is more in the research and academic arena. So I think we're still in an environment where the system sales are more oriented toward deeper research investigations where you need control over more of the operating parameters of the system than what's required if you're just trying to run a commercial application.

But the reason why I say I think we may see one commercial purchase of an enabling quantum computer is because I think at least in the blockchain and AI arena, we may see commercial organizations with an interest in doing some research into how these systems ultimately will be able to benefit AI and/or blockchain. So mostly still research and academically oriented, still purchasing systems to be able to control more of the operating parameters that you can't control when you're running on a cloud-based service. But possibly one commercial sale this year in an emerging application area where there's some research to be done that will require more control over the parameters of the system.

Now I think that what is likely the potential to change in a significant way, system sale purchase from research and academic to commercial is if we're successful with the work that we're doing on blockchain and AI. I think that those 2 areas could potentially be very transformative to D-Wave with respect to significant commercial sales of systems in support of those application areas. But we're not there yet. We're making good progress. I've talked about this in the past. I think the launch of the Test Net with post-quant labs for a new quantum blockchain -- classical blockchain environment is a really good next step -- and we're hopeful with respect to what we will see coming out of that work and potentially validating the application opportunity for our systems in that arena, but we're not quite there yet. I also think that not only the work we're doing with Shionogi on AI, but some other companies where we're now doing very similar work to what we've done with Shionogi could potentially help with that transition in AI as well. But in both of those cases, we're not quite there yet, but making good progress.

The next question comes from Craig Ellis with B. Riley Securities.

Congratulations on upping the system shipment outlook, guys. I wanted to ask a question on the nice detail you provided with QCI road map. The question is that what are you hearing from your 100 commercial customers on where they want to engage with that road map system capabilities? Is it at the 2028 level, 2030, 2032 level? And to what extent are you seeing QCI start to be additive to your potential customer base?

Sure, Craig. So we actually have a handful of customers that have expressed interest in the gate model system today. A couple have expressed interest in acquiring a gate model system and a few accessing it over the cloud. We are working on moving the tools into our Leap cloud service, integrating them with our Ocean SDK. And we are working on moving the actual hardware into the Leap Cloud service as well for cloud-based access to the system. We're also working on hardening the systems so that we could support sales of a gate model system, premise-based installations of a gate model system.

There's an understanding of the fact that the current system that we have operational is only 8 qubits -- but as we said in the past, we expect to have 17 qubits operational before the end of this year. And honestly, for both the cloud-based access and premise installation, there's interest in either the 8 or the 17 qubit system. In other words, we're not hearing -- come back to us when you've got a 49 or 175 qubit system. We're interested in getting our hands on these things now. So I think we may start to see some preliminary sales this year, but more likely into 2027.

The next question comes from Krish Sankar with TD Cowen and Company.

A two-part question. John, can you give some color on the composition of the backlog on RPO? How much is commercial, et cetera? And Alan, thanks for the color on the commercial adoption. I'm curious like some of these research academic sales that you're doing, are these niche R&D projects? Or can some of those lessons learned be ported over to accelerate commercial adoption?

So John, would be fair to everybody, you go ahead and answer the first question and glad to defer the second question because we did say only one question per.

Sure. So with respect to the makeup of the backlog, so we have $20 million of that is the system sale to FAU. And then we also have a very significant portion of the commercial enterprise SaaS deal that we did. So that backlog is roughly 50-50 between commercial and research.

Chris, feel free to get back in the queue.

The next question comes from Tyler Anderson with Craig-Hallum.

This is Tyler on for Richard Shannon. Have you gotten your hands on any of your multichip processors? And if so, what's the initial read and learnings from those? And any comment on coherence time would be helpful. And if you haven't, just any time line you would expect to would be great.

Okay. So are you talking about our annealing multichip processor?

Either one of them, -- whatever you have already going out.

Yes. So there are two things that we are working on for the multichip processors. And we've talked about it in the past, we are making good progress. One is obviously the bonding process between the processor chips and the other is scalable I/O. We are quite unique in the superconducting quantum computing arena in that we're controlling 4,000 qubits with 200 I/O lines versus everybody else. It requires 3 to 5 I/O lines per qubit. And that's due to our chip cryogenic control capability. However, as we are scaling from 4,000 annealing qubits to ultimately, for its 300,000 annealing qubits, that I/O needs to change.

The architecture needs to be a bit more scalable than it is currently. And so we now have masks and chips back that represent both the interconnecting of the processors as well as the new scalable I/O architecture. And so we're about to begin testing of those chips. So this is very much an R&D work in process. We're making good progress. We've defined the new scalable I/O architecture. We've created the initial masks to build out that capability. We've got early prototype chips back that we're going to begin testing. And we're in a similar position on the processor -- the bonding of the processor chips.

The next question comes from Ruben Roy with Stifel.

You could maybe a rough idea on the split between sort of how to think about upfront revenue, installation calibration, et cetera multiyear. And just to add on to that is on the RPO, CRPO for 12 months, some of the installation. Is that correct? [Technical Difficulty] Yes. I was just asking on the FAU system sales. If you can give us a rough idea on the split between initial installation versus sort of multiyear service and additional components to that sale? And then can you tell us about the cRPO, so the 12-month RPO, does that include some of the system sales to FAU?

The answer to your second question is, yes, the RPO includes FAU. And I cannot provide you detail on the elements of the rev rec on that system yet.

The next question comes from Troy Jensen with Cantor Fitzgerald. [Operation Instructions]

On the bookings and the momentum here. Just for you, Alan, I'd like to hear your thoughts on the NVIDIA announcement, the icing. Is this less important to D-Wave given the dual rail technology you guys have and obviously better fidelity so less need for air correction.

Yes. So first of all, I do want to comment on their use of the term icing. Anealing quantum computing is basically based on the icing Hamiltonian. And so typically, when we talk about programming the annealing quantum computer for the technical folks, we talk about converting your problem either into a Qubodraatic unconstrained binary optimization problem or an icing model problem. The two are equivalent. One is computer science speak, the other is physicist speak. However, the announcement from NVIDIA with use of the term icing has absolutely nothing to do with the icing amamiltonian or the icing programming model for annealing-based quantum computers. I'm not entirely sure why they picked that name.

But basically, the work that they are focused on with respect to leveraging GPU technology to aid in error correction is important work. I mean, there is a significant classical component to error correction. This is something that a lot of people don't really think much about or focus on. And in fact, it is one of the things that really makes solving optimization problems on gate model systems very inefficient. That classical overhead associated with error correction eats up pretty much all of the benefit of solving the optimization problem on a gate model system, whereas annealing quantum computers don't have that issue.

But nonetheless, there is a significant classical component to error correction. GPUs are an important component of the computing landscape for performing that piece of the computation. So in the context of our gate model work, what NVIDIA is talking about is absolutely relevant. That having been said, error correction on a dual rail processor is quite different from error correction on kind of standard, older technology qubits. The error correction is far more sophisticated and far more efficient. GPUs are still going to be important, but the combination will be done in a slightly different way. And so the work that NVIDIA is doing is relevant, but not quite as directly applicable to us in our dual rail technology. There's modification that will be required.

The next question is a follow-up from Joe McCormick with Evercore.

A quick one for John. Maybe, John, can you explain the deferred revenues dynamics? Because I think you had mentioned it's included in the RPO and stepped up a little bit. And I believe that was kind of related to Quantum Circuits, but maybe just to hear kind of the step-up there on deferred revs and how to think about kind of the evolution of deferred revs and how it will impact your backlog moving forward as well?

Well, deferred revenue is one of the components of the RPO number. But I can't provide that to in terms of specific accounts, but it is one of the components of the $43 million in backlog.

And we have a follow-up from Krish Sankar with TD Cowen.

Alex, I had a question for you. Your research academic sales, are these for niche R&D projects? Or do you think lessons learned there can actually be imparted to advancing commercial adoption faster?

Yes. So -- some of the research that's going on is more pure science research. There's been some interesting work recently out of Google that was kind of pure science work. And we've got researchers that are leveraging our systems similarly to do some pure scientific research, basically investigating physics theories that up until now have not been demonstrated or analytically validated. And so this is very important work, very interesting work. but not necessarily commercially application relevant.

Then there's other work that is more commercially relevant. For example, last year, we sold a system to the ULix Supercomputing Center. That system was installed -- delivered and installed last year. It's in their hands now. They are interconnecting it to their Jupiter Exascale supercomputer, 25,000 NVIDIA GPU system for work on optimization -- new optimization and AI workflows. And so the work coming out of that will absolutely be commercially relevant. I think the work that we will see at Florida Atlantic University is also exploring more commercially relevant application areas. And so I think it's a mix.

And we have a follow-up from Tyler Anderson from Craig-Hallum.

So with the blockchain application, is there anything specific about this blockchain that makes it amenable to your system? Are you able to address all proof of work protocols as well as proof of stake? Or is there a subset? Just like to know what makes this work.

So no, this is a new proof of work protocol that is by construction, quantum safe. And if you're doing the mining on the quantum processor, we believe will be much more energy efficient. However, as I said, we are about to enter a benchmarking phase within the test net to really understand the accuracy of the statement that I just made. So for now, that statement about energy efficiency is a hypothesis, not a fact, and we are beginning the benchmarking work to validate or not that statement. But it is a new proof of work protocol.

What makes the quantum computer able to win the majority of the blocks right now is that at its core, the proof of work is drawing samples from a distribution. And the quantum computer is very, very fast and very, very energy efficient at generating samples from a distribution, whereas CPUs and GPUs have a much heavier lift and they're much slower. And so if the proof of work requires you to generate multiple samples, in theory, it gives an edge and potentially a very significant edge to the quantum processor. So this is not about performing the existing proof-of-work computations. This is about a brand-new proof-of-work computation that can be performed either classically or quantum. You can use CPUs, you can use GPUs, you can use the quantum processor, but it absolutely in theory, gives a significant edge to the quantum processor. And if that holds up, then basically, we would have an architecture that is quantum safe and much more energy efficient.

This concludes our question-and-answer session. I would like to turn the conference back over to Dr. Alan Barrett for any closing remarks.

Okay. So let me just close with this. The quantum computing shakeout is coming. The industry is moving from promise to proof and from experiments to evidence. We believe that D-Wave is exceptionally well positioned for that transition because we are already delivering results in the market today while continuing to build differentiated technology for the future. We are not trying to win a corner of quantum. We are building to win across the market. With Annealing, we're driving commercial value now. With gate model, we believe we have a highly differentiated path to long-term leadership. And across both, we are making quantum computing easier for customers to adopt, easier to use and easier to generate value. RE is not waiting for the future of quantum computing. We are helping to define it now. So thanks again for joining us today, and we look forward to continuing the conversation at our Investor Day on June 1. We'll see you there. Thank you.

The conference has now concluded. Thank you for attending today's presentation. You may now disconnect.

Good morning, everyone, and welcome to D-Wave's Fourth Quarter Fiscal Year 2025 Earnings Conference Call. Today's conference call is being recorded. At this time, I would like to turn the call over to Kevin Hunt, Senior Director of Investor Relations. Please go ahead.

Thank you, and good morning. With me today are Dr. Alan Baratz, our Chief Executive Officer; and John Markovich, our Chief Financial Officer. Before we begin, I would like to remind everyone that this call will contain forward-looking statements, which are subject to risks and uncertainties and should be considered in conjunction with cautionary statements contained in our earnings release and the company's most recent periodic SEC reports. An on-demand webcast will be available, and a transcript of the conference will be posted on the Investor Relations section of the website within 48 hours after the call.

During today's call, management will provide certain information that will constitute non-GAAP financial measures under SEC rules, such as non-GAAP gross profit, non-GAAP gross margin, adjusted EBITDA loss, adjusted net loss and adjusted net loss per share and operating metrics such as bookings. Reconciliations to GAAP financial measures, definitions and certain additional information are also included in today's earnings release, which is available in the Investor Relations section of our company website at www.dwavequantum.com.

Given that D-Wave is now fortunate to have 15 sell-side security analysts publishing research on the company, we'd be limited to taking one question from each analyst during the first round of questions and then, time permitting, proceed to a second round of questions where again, we will have time to -- or have to limit each analyst to one question. I'll now hand over the call to Alan.

Good morning, and thank you all for joining us today. Fiscal 2025 was not just a strong year for D-Wave. It was an inflection point for the company and for the quantum computing industry. For years, this sector has been defined by unrealized promises, dependence on government grants and an inability to deliver customer value. In 2025, D-Wave separated itself from that narrative. We delivered proof, we delivered revenue, and we delivered real-world advantage.

While 2025 was declared the international year of quantum science and technology, for D-Wave, there was something more important. The year quantum computing moved decisively from research to real-world impact. And we believe that no company advanced that transition more than D-Wave. 2025 was a year of objective proof, evidence of D-Wave's technical and commercial progress. We began the year by closing our first Advantage quantum computer system sale to the Julich Supercomputing Centre, marking the first time a commercial annealing quantum computer was purchased for integration into a national supercomputing facility.

We then became and remain the only quantum computing company to demonstrate quantum supremacy on a useful real-world problem. That result, which was achieved natively on our Advantage2 quantum processing unit has not been successfully challenged for nearly 2 years after the paper's initial publication. This demonstration was an entirely quantum computation, not a hybrid computation. Moreover, no other companies other than D-Wave Google and Quantinuum have achieved quantum supremacy on any problem, not IBM, not IonQ, not Rigetti, not Infleqtion, not Xanadu, not IQM. All attempts other than D-Wave, Google and Quantinuum have been spoofed. And only D-Wave's result was on a useful real-world problem.

Then in May, we launched general availability of our Advantage2 system, the same system that achieved that supremacy milestone. And critically, we transitioned that technical leadership into commercial performance with record revenue of $24.6 million in fiscal 2025, up 179% year-over-year, $13.4 million in Q4 bookings, the second highest quarterly bookings in the company's history and up 471% from the immediately preceding third quarter and a sales opportunity pipeline that expanded by nearly 1,500% year-over-year.

In an industry long on promises, D-Wave is delivering measurable results. And now we have entered 2026 with extraordinary momentum. In January alone, we generated more bookings than in the entirety of fiscal 2025. We closed a $20 million system sale with Florida Atlantic University. We signed a 2-year $10 million enterprise Quantum Compute as a Service agreement with a Fortune 100 company, one of the largest enterprise Quantum Compute as a Service deals in the history of the quantum computing industry, and we completed the acquisition of Quantum Circuits.

The acquisition of Quantum Circuits fundamentally changes the competitive landscape. We believe that it firmly secures D-Wave's position as the world's leading quantum computing company and the only dual platform quantum computing company. As a result, we believe that we are the only company positioned to address the full quantum computing market. Our dual platform approach is important because it allows D-Wave to be a one-stop shop capable of solving the full range of the complex problems that customers face.

But let me be clear, this approach is not new. It has been our strategy for 5 years. We dominate optimization today with annealing quantum computing technology. And now by combining Quantum Circuits' industry-leading dual-rail qubit technology with D-Wave's proprietary on-chip cryogenic control, we are also positioned to be the leader in error-corrected gate-model systems.

Annealing quantum computing remains a strategic focus for D-Wave. Optimization is one of the largest and most immediate commercial opportunities in quantum computing. It spans logistics, defense, telecom, manufacturing, finance and energy, virtually every major industry and optimization problems require annealing quantum computing. D-Wave has demonstrated material performance advantages over classical approaches across a multitude of optimization use cases. We are running production workloads today. To our knowledge, no gate-model quantum computing company has demonstrated a practical advantage over classical systems for optimization. And academic literature suggests they likely never will.

We're also seeing early promise with customers exploring annealing quantum computing's impact on AI and blockchain, 2 areas with enormous commercial potential. Annealing is not a stepping stone. It's a commercially proven architecture with expanding performance gains. And with our Advantage3 system in development, we expect to further extend that performance gap. Annealing dominates optimization today, and we believe that it will continue to dominate as the market expands.

Now let's talk about gate-model. Most superconducting competitors are pursuing legacy transmon architectures that require massive physical qubit overhead for effective error correction and complex wiring schemes that will struggle to scale economically. With Quantum Circuits, D-Wave takes a different path. The Quantum Circuits deal is transformational. With it, we believe that D-Wave gains a decisive architectural advantage. Dr. Rob Schoelkopf, the inventor of the transmon qubit used by nearly every superconducting competitor moved beyond that architecture to develop dual rail qubits with built-in erasure detection that identifies 90% of errors that occur. With erasure detection, this technology delivers gate fidelities that exceed 99.9%, bringing trapped ion fidelities along with superconducting execution speeds to today's gate-model algorithm developers.

Our eraser detection and our observed erasure rate of 0.5% allow us to deliver logical qubits with an order of magnitude fewer physical qubits compared to architectures without this capability. Error correction is essential to unlocking broad quantum utility, and we believe that the dual rail technology offers the fastest path to large-scale error corrected architectures. I can't emphasize this enough. The dual rail technology allows us to achieve superconducting speed with the fidelity of ion trap or neutral atom approaches. This is an industry game changer unmatched by any other gate-model vendor today.

The implications of D-Wave's dual rail technology are profound. Our approach achieves logical qubit ratios of roughly 1 logical qubit for every 100 to 200 physical qubits compared to about 1 logical qubit for every 1,000 to 2,000 physical qubits in conventional superconducting designs. What's equally remarkable are the gate speeds. Dual rail gate speeds are 1,000x faster than ion trap or neutral atom systems. The fidelity of ion trap or neutral atom approaches with the speed of superconducting. That's a fundamental improvement in the metrics that matter. Speed matters, error correction overhead matters, scalability matters and D-Wave now holds advantages in each.

But our gate-model innovations don't stop there. In January, D-Wave demonstrated that the on-chip cryogenic control currently being used in its advantaged quantum computers can be used to control gate-model qubits without loss of fidelity. This industry-first milestone advances the development of commercially viable gate-model quantum computers by providing a path to significantly reduce the wiring required to control large numbers of qubits. We are now working on leveraging this technology to provide full qubit control at scale. This would ultimately enable the ability to control gate-model systems with multiple orders of magnitude fewer control lines than required by competing superconducting gate-model systems. That difference is not incremental. It's architectural, it's essential.

As we discussed at the time of the Quantum Circuits acquisition, we have an 8-qubit gate-model system available to select customers today, and we expect a 17-qubit system later in 2026. We've already seen tremendous interest from customers and expect to start generating some Quantum Compute as a Service revenue from our gate-model systems this year while also building a pipeline of gate-model system sales opportunities for delivery beginning in 2027. So in addition to accelerating commercial momentum that we see with our Advantage2 annealing system, we expect our gate-model offering to deliver a small but growing stream of revenue in 2026.

We believe that the Quantum Circuits' acquisition positions D-Wave as the leading contender to deliver the first fully error corrected scalable superconducting gate-model quantum computer. This effectively doubles our long-term addressable market by delivering both annealing and gate-model quantum computing solutions. What's also particularly noteworthy is our rapidly accelerating commercial traction, which reflects a differentiated strategy from most all other quantum computing companies. At our qubits conference in January, our largest and most successful user conference ever, we announced a $20 million Advantage2 system sale with Florida Atlantic University as well as a 2-year $10 million enterprise QCaaS agreement with a Fortune 100 company, one of the most significant enterprise deals in the history of the quantum computing industry. This is not research revenue. It represents commercial adoption by a growing collection of the world's largest companies.

D-Wave is currently engaged with some of the world's leading airlines, payment companies, telecom operators, defense contractors, chemical companies, health care companies, aerospace companies and more. Customers are no longer asking if Quantum will be useful someday. They are asking how quickly they can deploy it. The U.S. government is also taking note. To support escalating government interest in both our annealing and gate-model technologies, we recently launched a dedicated U.S. Government Solutions business unit. Unlike other quantum companies that are focused primarily on securing federal R&D grants and characterizing those as commercial revenue, our strategy is very straightforward, solve real mission-critical problems now and derive government revenue today.

At Qubit, we demonstrated a missile defense simulation in collaboration with Davidson Technologies and Anduril. For a 500 missile attack simulation, we showed a 10x faster time to solution, a 9% to 12% improvement in threat mitigation and a 45 to 60 additional missile intercepts. As complexity increased, D-Wave's technological advantage increased. This is operational relevance.

Anduril's President and Chief Business Officer, Matthew Steckman, spoke during my Qubit's keynote and indicated that he was surprised at how fast and mature D-Wave's technology is. And he suggested that there are a lot of potential for collaboration going forward. We believe that there is significant opportunity in U.S. government applications across both our annealing and gate-model platforms.

We're also seeing growing interest in system sales. In addition to annealing quantum computing system-related agreements with Julich in Germany, Q-Alliance in Italy and Florida Atlantic University in the U.S., we continue to advance discussions in South Korea as well as with additional HPC, academic and government institutions.

On the gate-model side, we expect to see the development of a multimillion dollar R&D system sales pipeline for 2027. For both our annealing and gate-model computers, these are premium-priced systems with high gross margin profiles. Underpinning all of D-Wave's technical and commercial traction is a very strong leadership team with decades of deep expertise in their respective areas of focus.

We recently brought on Jack Sears, Jr. to lead U.S. Government Solutions, Stan Black as our Chief Information Security Officer; and as I previously mentioned, Dr. Rob Schoelkopf, who brings world-class superconducting leadership and maintains strong ties with Yale University.

Our Chief Development Officer, Dr. Trevor Lanting, will oversee product development across both annealing and gate systems, ensuring integration, speed and execution. The strength of our management team and its track record of success are key to maximizing D-Wave's near-term opportunities and long-term growth.

Our operational footprint and workforce also continue to expand with the announcement that D-Wave's headquarters will relocate from Palo Alto, California to Boca Raton, Florida later this year, where we will also open a major U.S.-based R&D center. With this expansion, D-Wave will operate 3 main R&D hubs: Burnaby, British Columbia, New Haven, Connecticut and Boca Raton, Florida. We are building a distributed innovation footprint designed to attract top-tier quantum talent, provide bicoastal redundancy in case of disaster recovery and ultimately lead the next era of computing.

Let me close with a broader industry observation. Quantum computing is entering a new phase. The first phase was scientific exploration. The second phase was capital formation. The next phase will be commercial separation. Over the next several years, we expect that this industry will consolidate around a small number of companies that can demonstrate 3 things: real performance advantage, real commercial adoption and a scalable, economically viable architecture. Many will not make that transition. D-Wave already has.

We are the only company to demonstrate real-world quantums of primacy on a useful problem. We are the only company running production applications for Forbes Global 2000 enterprise customers. We are the only dual platform quantum computing company with a commercially proven annealing quantum computer generating meaningful revenue and a differentiated superconducting gate-model platform with a credible pathway to full error correction.

Others are still pursuing proof of concept, we have proof of commercialization. Others are dependent on long development time lines, government funding and sustained capital market support. We're building a business with commercial customers, contracts and expanding bookings. Others have made product development decisions that focus on either superconducting speed or ion trap and neutral atom fidelity. We can deliver both. As the market matures, capital will flow toward companies with operating leverage, commercial validation and technical defensibility. We believe D-Wave is uniquely positioned at that intersection.

The quantum industry will not support dozens of long-term winners. It will support a handful of durable platforms, and we intend to be one of them. Fiscal 2025 marked the moment when D-Wave moved from participant to frontrunner. The momentum we are seeing in early 2026 suggests that this gap is widening. With that, I'll hand the call over to John to provide a review of our fourth quarter and fiscal 2025 results. John?

Thank you, Alan, and thank you to everyone for taking the time to participate in today's call. In my review of the fiscal year '25 and fourth quarter results, I will be providing non-GAAP operating metrics, including bookings as well as non-GAAP financial measures that include non-GAAP gross profit, non-GAAP gross margin, adjusted net loss, adjusted net loss per share and adjusted EBITDA loss as we believe these metrics improve investors' ability to evaluate our underlying operating performance. These measures are defined in the tables at the bottom of today's earnings press release with the non-GAAP financial measures for the most part adjusting for noncash and nonrecurring expenses.

Revenue for fiscal 2025 totaled $24.6 million, an increase of $15.8 million or 179% from fiscal 2024 revenue of $8.8 million with fiscal '25 revenue including $16.2 million in systems sales revenue, $5.5 million in QCaaS subscription revenue and $2.7 million in professional services revenue. I would like to highlight several aspects of D-Wave's revenue that clearly distinguishes the company from a number of other so-called quantum computing companies.

First, all of our revenue is derived from selling, providing access to or providing services for quantum computing systems. We do not recognize any revenue from any products or services that are not directly related to quantum computing, such as quantum sensing, quantum networking or encryption systems that rely on quantum physics, but not on quantum computing. None of these products or services have anything to do with quantum computing that we define as computing systems that harness quantum mechanical effects, specifically super position and entanglement to solve complex computational problems. In addition, we do not give, grant, invest or lend funds to any of our customers that they utilize or intend to utilize towards the purchase of our products and/or services.

Fiscal 2025 bookings were $18.7 million, a decrease of 22% or $5.2 million from fiscal '24 bookings of $23.9 million, keeping in mind that the 2024 bookings included an 8-figure booking of the company's first system sale. Subsequent to the end of fiscal 2025, D-Wave has closed over $32.8 million in additional bookings that includes a $20 million system sale to Florida Atlantic University and a $10 million 2-year enterprise license deal with a Fortune 100 company.

With respect to the diversity of our customer base, in fiscal 2025 D-Wave recognized revenue from over 135 individual customers, encompassing over 70 commercial customers that includes over 2 dozen Forbes Global 2000 enterprises. The average revenue per commercial customer increased by 20% over fiscal '24 and the total revenue recognized from Forbes Global 2000 customers increased by 70% on a year-over-year basis with the average Forbes Global 2000 deal size up 90% on a year-over-year basis.

GAAP gross profit for fiscal '25 was $20.3 million, an increase of $14.7 million or 265% from fiscal '24 GAAP gross profit of $5.6 million with the increase due primarily to a higher-margin quantum computer system during the year. Non-GAAP gross profit for fiscal '25 was $21.1 million, an increase of $14.7 million or 229% from the prior year non-GAAP gross profit of $6.4 million.

GAAP gross margin for fiscal '25 was 82.6%, an increase of 19.6% from fiscal '24 GAAP gross margin of 63%, with the increase again due primarily to a higher-margin quantum computer system sale during the year. Fiscal '25 non-GAAP gross margin was 86%, an increase of 13.2% from the prior year non-GAAP gross margin of 72.8%. Again, the difference between GAAP and non-GAAP gross profit and gross margin is limited to noncash stock-based compensation and depreciation and amortization expenses that are excluded from the non-GAAP gross profit and gross margin measures.

Net loss for fiscal 2025 was $355 million or $1.11 per share compared with the fiscal '24 loss of $143.9 million or $0.75 per share. The increase in net loss was primarily driven by $250.5 million in noncash nonoperating charges related to the remeasurement of the company's warrant liability as well as realized losses stemming from warrant exercises, both directly related to the increase of the price of the company's warrants and common stock. Excluding this noncash remeasurement charge, the adjusted net loss for fiscal '25 was $84.5 million or $0.26 per share, an increase of $8.9 million or 11.8% when compared to the fiscal '24 adjusted net loss of $75.6 million or $0.39 per share. The reduction in net loss per share was due to a higher issued and outstanding number of common shares in '25 when compared to '24.

Adjusted EBITDA loss for fiscal '25 was $71.8 million, an increase of $15.8 million or 28% from the fiscal '24 adjusted EBITDA loss of $56 million, with the increased loss due primarily to higher operating expenses, partially offset by higher gross profit.

Now I will move on to the fourth quarter. Revenue in the fourth quarter totaled $2.8 million, an increase of approximately $0.5 million or 19% from the fourth quarter of fiscal '24 revenue of $2.3 million, with fourth quarter '25 revenue including $1 million in QCaaS subscription revenue, $1 million in professional services revenue and approximately $700,000 in systems sales revenue.

Bookings for the fourth quarter were $13.4 million, a decrease of $4.9 million or 27% when compared to the year earlier quarter of $18.3 million that included the 8-figure system sale that I referenced earlier. On a sequential quarter-to-quarter basis, bookings increased $11 million or 471% from the immediately preceding fiscal '25 third quarter bookings of $2.4 million, with the increase due primarily to the previously announced EUR 10 million booking for a multiyear 15% -- 50% capacity commitment for a D-Wave Advantage to annealing quantum computing system to support the development of a Lombardy, Italy-based state-of-the-art quantum computing and research facility.

GAAP gross profit for fiscal '24-'25 fourth quarter was $1.8 million, an increase of approximately $300,000 or 21% from the fiscal '24 fourth quarter gross profit of $1.5 million, with the increase due primarily to the growth in revenue. For the fourth quarter, non-GAAP gross profit was $2 million, an increase of approximately $300,000 or 17% from the prior year fourth quarter non-GAAP gross profit of $1.7 million.

GAAP gross margin for the fiscal '25 fourth quarter was 64.8%, an increase of 1% from the fiscal '24 fourth quarter GAAP gross profit margin of 63.8%. For the fourth quarter, the non-GAAP gross margin was 71.8%, a decrease of 1.2% from the fiscal '24 fourth quarter non-GAAP gross margin of 73%.

Net loss for the fourth quarter of fiscal '25 was $42.3 million or $0.12 per share, a decrease of $43.8 million or $0.25 per share from the fiscal '24 fourth quarter net loss of $86.1 million or $0.37 per share. The decrease in net loss was primarily due to a decrease of $57.7 million in noncash nonoperating charges related to the remeasurement of the company's warrant liability, partially offset by higher operating expenses. Excluding this charge, the fourth quarter adjusted net loss was $31.8 million or $0.09 per share, an increase of $14 million or $0.01 per share from the fiscal '24 fourth quarter adjusted net loss of $17.8 million or $0.08 per share.

Adjusted EBITDA loss for the fourth quarter was $25 million, an increase of $9.7 million or 63% for the prior year fourth quarter adjusted EBITDA loss of $15.3 million, with the increase due primarily to higher operating expenses, partially offset by higher gross profit.

Now I will address the balance sheet and liquidity. During fiscal '25, D-Wave raised over $800 million in gross proceeds from the issuance of equity under 2 ATM programs, an ELOC program and from the exercise of warrants and stock options. As of December 31, 2025, D-Wave's consolidated cash and marketable securities balance totaled $884.5 million, representing a 397% from the year earlier consolidated cash balance of $178 million and a 6% increase from the immediately prior fiscal '25 third quarter consolidated cash balance of $836.2 million. During the fourth quarter, the company received $63.7 million in cash proceeds from the exercise of warrants.

As previously announced, subsequent to year-end, we invested $250 million in cash in conjunction with the acquisition of Quantum Circuits, and we believe that our remaining liquidity is sufficient to support a fully funded plan to profitability.

With respect to 2026, we will continue our practice of not providing formal financial guidance. However, I would like to provide some parameters. With respect to bookings, we are obviously off to a tremendous start with fiscal 2026 year-to-date bookings already exceeding our annual bookings for any year in the company's history.

As Alan noted earlier, our sales opportunity pipeline entering 2026 was up nearly 1,500% to the beginning of 2025 that includes a 700% increase in the total number of prospective sales transactions. And we continue to see interest in potential system sales, not only for our Advantage2 annealing system, but also for our dual rail gate-model quantum systems. However, as we have previously noted, the system sales process is fairly complex and the sales cycle is usually lengthy in duration.

With respect to revenue recognition, on system sales, please keep in mind that most of these transactions will involve site preparation, installation, calibration and other key steps before the systems are fully operational. They are likely to encompass multiple months and possibly quarters depending on the unique elements of a particular system transaction. As a result, our revenue recognition on system sales is on a percentage of completion basis.

In addition, we anticipate that most system sales transactions will involve a multiyear service and maintenance revenue component and some may include a multiyear Leap cloud access component.

With respect to the recently announced $10 million enterprise QCaaS agreement, this revenue will be recognized ratably over a 2-year time frame, commencing in the current quarter. The EUR 10 million booking in Italy will be recognized ratably over 5 years, commencing once the system is fully installed, which we expect will be in the second half of this year. To summarize, we expect incrementally higher revenue growth in the second half of this year when compared to the first half.

With respect to operating expenses, we intend to continue to invest aggressively in both our annealing and gate-model technology development initiatives that consist primarily of research and development headcount, fabrication expenses and to some degree, capital expenditures.

As we previously outlined, approximately 65 research and development professionals joined D-Wave through the Quantum Circuits acquisition, and we intend to expand this New Haven, Connecticut-based gate-model team by at least 50% over the course of this year. In addition, we will be making significant headcount and capital investments at our recently announced U.S. R&D facility in Boca Raton, Florida, where we intend to expand our annealing R&D team and eventually to install one or more annealing systems to support our Leap cloud service offering.

Lastly, given the recent formation of our government business unit, we will be making meaningful investments in this area given the magnitude of opportunities that we see here. Over the course of fiscal '26, we expect to increase quarterly operating expenses by approximately 15% sequentially over the immediately prior fiscal quarter.

In conclusion, as we have previously stated, we continue to believe that D-Wave has the opportunity to be the first independent publicly held quantum computing company to achieve sustained profitability and to achieve this milestone with substantially less funding than required by other independent publicly held quantum computing companies.

With that, operator, please open the call for questions.

[Operator Instructions] Our first question comes from Harsh Kumar of Piper Sandler.

Congratulations on multiple fronts and all the progress you guys are making on technological achievements. Alan, I had one for you. I wanted to ask about the gate-model, particularly. Can you help us think about the advantage with the built-in error correction that you have? And I'm asking specifically from a time-to-market standpoint. Or should I just think of this as you just need a lot less, like 10x less qubits to get to a commercial system. But I've got to believe this translates to some kind of a timing advantage for you as well.

Yes. So yes, it absolutely does translate into a timing advantage. And the reason it translates into a timing advantage is because the complexity of building a fully error corrected system and scaling it to the size needed to achieve quantum utility is much lower, given the dual rail technology combined with our cryogenic control technology. So the dual rail technology gives us very high gate fidelities on par with some of the best in the industry, including trapped ion and neutral atom, but while preserving the 1,000x speed advantage of superconducting over the other modalities.

But because of the higher fidelities, we are able to error correct with many fewer physical qubits per logical qubit. So that is a complexity reduction advantage. And then when you add the on-chip cryogenic control and what we believe will ultimately allow us to control hundreds of thousands or more qubits with just hundreds of IO lines, versus hundreds of thousands of I/O lines as with other superconducting approaches, again, a dramatic reduction in the complexity. We believe that this combination will allow us to build and deploy scaled error-corrected superconducting gate-model systems ahead of anybody else. And we believe that, obviously, ultimately, superconducting will leave because of the speed advantage.

Our next question comes from Quinn Bolton of Needham & Co.

I'll also offer congratulations on a great 2025. I guess, John, I just wanted to come back. I know you're not giving guidance per se for '26. But as we think about the integration of QCI now into the business, can you give us any sort of points as to how much OpEx you would expect to incur in 2026 with that acquisition? Is the OpEx for that team sort of fully factored into that 15% OpEx quarter-on-quarter increase that you talked about for March? Just trying to think about if we had a base model for D-Wave prior to the acquisition, how much additional OpEx would you think we would be putting into the models for the year now that the QCI is closed?

Quinn, my comments earlier in terms of the sequential 15% growth in OpEx, that includes QCI expansion in not only their R&D team, but other expenses, including fabrication expenses as well as some capital expenditure. So the answer to your question is yes, that 15% fully includes the incremental costs associated with QCI.

Our next question comes from Kingsley Crane of Canaccord.

Congrats on the really strong momentum. John, for you, QCaaS, it's moderated a bit the past year with system sales driving growth. I'm just wondering if you have a sense of what you think an ideal net retention rate for that segment can be or what that could be this next year? And then if Stride hybrid solvers' new ML integration capabilities could change that upsell conversation at all with existing customers?

Kingsley, we have not published retention rates. But what I can tell you is when you take a look at the composition of the QCaaS over the last 2 years, keep in mind that a significant component of our QCaaS in '24 was Julich, and that transitioned to a system sale. And when you take a look at the bookings that we recently announced, including the 50% utilization of the system in Italy as well as the enterprise QCaaS deal, we're starting to see substantially larger transactions that give us incrementally more visibility on the growth in the overall QCaaS than what we've seen in the past.

And John, maybe I'll just chime in as well, especially given the comment about the Stride solver. So Kingsley, we've been saying all along that system sales and QCaaS are very complementary models for us and that the system sales for now are larger deals and nearer-term revenue recognition versus the QCaaS deals, which are a bit smaller and recognized ratably over a multiyear period in general. However, we are really beginning to see an increase in the size of QCaaS deals. I started signaling this last year when I said we're now looking at larger companies doing larger deals with us, including potentially enterprise all-you-can-eat licenses. And that 2-year $10 million Fortune 100 company deal that we closed at the beginning of this year is the first example of that. That is starting to transition QCaaS into larger enterprise license deals.

Now revenue for those deals still gets recognized ratably. So the revenue recognition is generally over a longer period of time than the system sales, but we're really now starting to see the growth in the size of those QCaaS deals. So I think that as we look to the future, we will continue to see larger system sales deals than QCaaS deals, earlier revenue recognition on the system sales than the QCaaS, but I think we're going to really start to see QCaaS picking up the pace as we aim to do more enterprise licenses.

Our next question comes from Joe McCormack of Evercore.

Thanks for taking the question on for Mark Lipacis. Maybe just around Quantum circuits and how it's playing into that kind of sales pipeline increase that you're talking to. I saw that you closed a little bit north of $2 million in bookings for QCI in January. And so can you speak a little bit to the levels of engagement that you're seeing? And I don't know if there's any kind of qualification around like what that book of business kind of looks like from a backlog perspective that's folding in as we enter the 2026 year?

John, you want me to do that or you want to take it?

No, I'm happy to. As we articulated when we first announced the transaction, Joe, we do expect revenue contribution over the course of this year from Quantum Circuits on the professional services and QCaaS side. They also have a book of business that is government related. And they actually had some revenue last year that was government related. So that's where we expect revenue contributions from Quantum Circuits this year. And then as we've previously outlined, we also expect to start to develop a sales pipeline over the course of this year for potential systems sales.

And the only other thing I'll say is we are seeing a lot of interest in the dual rail systems, including the 8 qubit system that we have operational today with some early customers using it and the 17 qubit system we expect later this year. So a number of our current annealing customers have expressed interest in that system in addition to the annealing system. So we're quite encouraged by the interest that we're seeing.

Our next question comes from Krish Sankar of TD Cowen.

Alan, I just wanted to find out, like, obviously, one of your competitors is buying one of your foundries. Kind of wondering how you're looking at risk mitigation? And also, does QCI use the same foundry as qubit? Or is this a different foundry?

Okay. So currently, the dual rail technology is not fabricated at SkyWater. And for our annealing technology, SkyWater does fabricate the wiring, but they do not fabricate the active components, the Josephson junctions, which is, in some sense, the most important and sensitive technology fabrication component from an IP perspective. The active components, we fabricate ourselves in our R&D facilities and SkyWater does the wiring. And on dual rail technology, SkyWater is not involved at all. So my view on the IonQ acquisition of SkyWater is that on the one hand they're saying all the right things relative to continuing to work together exactly as we have been, and we should not be concerned about anything changing as a result of this transaction. On the other hand, we are skeptical, and we are concerned. And so we are actively working on other sources of fab support for our systems.

Our next question comes from Kevin Garrigan of Jefferies.

Let me echo my congrats on all the progress. Hey, Alan, you talked a little bit about it in a previous question. But on the QCaaS side, I mean, how are customer conversations evolving? And what metrics are customers really focused on when evaluating Quantum as a Solution? I mean, is it all about kind of speed up time? Or is it convenience or just your quantum annealer is far better than anything out there?

Well, I mean, first of all, the annealing quantum systems are the only ones that can actually deliver any commercial value today on real-world problems. And they're the only ones that are used in production by customers today. No other quantum computer is capable of that level of computation and commercial ROI. And the way this evolves is that we basically engage a customer on an initial application, and we've gotten very good at being able to identify upfront whether the application that they're interested in or other applications that they're dealing with will benefit from our systems or not.

So now unlike in the past, when we start an initial application development with a customer, we have a very high degree of confidence that they're going to be able to see a very strong ROI. And then that is what allows us to basically validate for them the benefits and the value they can get from working with us. That then allows us to move more rapidly to getting that initial application in production. And then that's what generates interest in other applications.

So the Fortune 100 deal that we did, that started with a first application. They were blown away by the results that we achieved, including a dramatic improvement in their bottom line based on using this technology. And then they came back and said, okay, we've got quite a few other applications, we want an all-you-can-eat license. We are now starting to see some other large companies see similar benefits from the initial application and talking about similar kinds of engagements.

Our next question comes from Ruben Roy of Stifel.

Congrats, Alan and John, on the progress. These are probably a little bit longer term in nature questions. But wondering with your annealing customers, obviously, you have a lot of commercial customers on the annealing side. Have you started to have some conversations on potential longer-term road map opportunities with gate-model QCI computing with those success stories on the annealing side? And then the second part of that question is -- and again, it's probably pretty early here, but you've got the dual platform approach. Are there opportunities in your view to combine annealing and gate to come up with unique solutions. Again, you're the only compute company -- quantum compute company with both. So I'm wondering if there are opportunities longer term to have hybrid solutions or whatnot to even expand the TAM further?

Okay. So first of all, yes, a number of our annealing customers have approached us and said, look, we've got some other use cases here that we'd like to look at in the context of your gate-model technology. Our customers understand the difference between annealing and gate. They understand the types of problems that require annealing versus the types of problems that require gate. We've kind of educated them on that. They're pretty savvy on that. And so they recognize that they've got other problems that potentially could benefit from gate, and they have started to engage us on those discussions.

And then the only other thing I'll say relative to your second question, and I shouldn't because we said only one question per, but it's an easy one, not yet. There's some early evidence based on the fact that we've integrated some digital controls, read that as some gate operations into our annealing systems, and we're seeing some very interesting scientific results based on that. So maybe, but it's way too soon to be thinking of that as a viable commercial opportunity.

Our next question comes from Troy Jensen of Cantor Fitzgerald.

Congrats on all the momentum here. Alan, I totally agree with you. I guess my best takeaway from Qubit '26 was there's just dozens of customers out there that have kind of piloted programs and seem ready to kind of move forward. So my question on like it's multiple 8-figure enterprise QCaaS deals. Can you talk about like how much capacity do you guys have with your existing annealing computers and the time that may take to launch more if you need to ramp quickly?

Yes. We have plenty of capacity in our Leap quantum cloud service to support, I mean, minimum tens of enterprise deals. Our quantum computers are very capital efficient. Each quantum computer can support $25 million to $30 million of revenue per year. We've got 4 of them available today. So we've got plenty of revenue capacity for these kinds of deals. But deploying another system, the capital cost is only about a couple of million dollars. And the build time, I mean, once we have the componentry is like 3 to 4 months. So with some lead time, we have no problem deploying additional systems.

Our next question comes from Craig Ellis of B. Riley Securities.

Congratulations on the real strong execution. Alan, I wanted to ask you a higher-level question, and I'll rewind the clock a little bit. I think it was 3 quarters ago, you told us to expect increased R&D and go-to-market spend. And here we are now able to show proof that we've got annealing cryo control applicability to gate, and we start the year with, I think, around $45 million in trailing 4 or 5-month bookings, which is extremely robust. So the question is, is that signs of execution of what you were pointing to? Or were you expecting something else? And as we start the year, if you can give us any color on what you see in the pipeline on the system side and with that all-you-can-eat newer offering, it would be greatly appreciated.

Yes. So the short answer is yes, that our investments were designed to accelerate work on our Advantage3 system, which includes analog digital capability as well as multichip for scaling to 100,000 qubits and then to really start accelerating work on the gate-model system. One of the key elements that we uniquely were bringing to the table was on-chip cryogenic control.

And so on the R&D side, yes, the investments are playing out as we had planned. And on the go-to-market side, you said it. I mean, we're making really good progress. It's robust at this point in time. Our pipeline has grown significantly, and we're feeling quite good about what we can expect to see this year. So the investments in go-to-market are playing out exactly as we expected as well.

Our next question comes from John McPeake of Rosenblatt Securities.

Great work. I like what you're doing. A question on Advantage3. Do you have any more information about circuit tests? Any information that you could give us relative to how you're progressing there and what the capabilities might be relative to Advantage2?

Yes. So I kind of called out the 2 key elements, one analog digital, the other multichip. Obviously, with each generation of system, it's more qubits, more connectivity and higher coherence times. But functionally, the big things for Advantage3 are putting some digital controls into the annealing fabric as well as multichip for scaling far more rapidly. We've got our first chips back that incorporate the analog digital controls, and we are close to having our first chips that demonstrate multichip interconnect. So we're making good progress on all fronts.

Our next question comes from David Williams of Benchmark.

Let me also echo my congrats on the execution here. Maybe, Alan, can you speak to some of the pipeline that you talked about in the script, just the strength there, where that's coming from and really what you're hearing from customers? And how quickly can this pipeline turn into maybe confirmed revenue or those orders can come in, just kind of that life cycle of that pipeline.

Yes. I'm not going to address the revenue piece because that's all based on the revenue recognition policies of the company and different deals have different recognition time lines. But as far as closing the deals, I mean, we've got a strong pipeline for both system sales. I mean, honestly, when we talked about this at the beginning of last year, I said expect maybe one a year for the foreseeable future. Our pipeline for system sales right now is very robust. So we're beyond 1 a year at this point in time.

And then the same is kind of true on QCaaS and professional services deals. I mean, I talked a little bit about it. One of the world's largest airline companies, one of the world's largest health care companies, largest chemical companies. I mean, we are closing deals with much larger companies. These are much larger deals from the outset, and we're progressing through them much faster. So very strong go-to-market environment for us right now.

Our next question comes from Antoine Legault of Wedbush Securities.

I also want to echo my congratulations on the progress in 2025. So there's been reports that Pentagon's budget would increase significantly into fiscal 2027. And with some of that budget likely to be allocated to quantum technologies. Can you tell us a bit more about the magnitude of the opportunity ahead and how that might benefit you, particularly given your new government business unit?

Yes. So just quickly, first of all we are not primarily pursuing R&D research grants. This isn't about the government funding us to build our systems. We've got plenty of liquidity to be able to fund our R&D road map. What we are focused on is helping the government solve their hard computational problems today. And in fact, we've got a very interesting pipeline there. I'll be very frank with you. When we talked about the Davidson Anduril deal at Qubits and Anduril talked about what they had seen in using our system, that generated a very significant inflow of interest in leveraging our systems to solve hard problems within the U.S. government. So we're feeling like we've got the wind to our back right now. And with Jack Sears on board and building that -- the government business, basically having a place where we can engage and deliver for the U.S. government, we're feeling quite good about the opportunity.

Our next question comes from Richard Shannon of Craig-Hallum.

This is Tyler Anderson on for Richard. So you mentioned that you have 50% of the capacity of your system booked. When we're thinking about future new systems that are coming online for the Advantage3 and beyond, is there a potential where we see multiple of those systems come online right away? And that way you can have that capacity reserved for customers that you're talking to today? Are you having those conversations? Just want to get some color on that.

Yes. So Tyler, first of all, when you say 50%, the only time we've talked about 50% of capacity was in the Q-Alliance deal in Italy. They purchased 50% capacity of an Advantage2 system. In general, in our cloud service today, we're not yet even at 50% capacity. We've got plenty of capacity in our Leap cloud service to support our professional services engagements and Quantum Compute as a Service customers.

When we bring a new system to market, we try to upgrade all our cloud systems as quickly as possible. And in the past, because the numbers have been relatively small, we've just done them one at a time. However, now that we are seeing a lot more interest in system sales, we are making some investment in the team and the capabilities to do installations so that we can -- so that we don't have to serialize. We can do more in parallel going forward. So the answer is yes.

Our next question comes from David Liu of Mizu.

I'm on for Vijay. Congrats on the strong momentum here. My question I wanted to ask is you guys called out the interest and momentum in system sales growth going into '26 as well as the enterprise traction for QCaaS. So how should we think about QCaaS and the hardware sales mix going forward? And maybe in relation to that, the OpEx number as well for the year?

Joh, do you want to take that?

Sure. With respect to the OpEx number, as I articulated earlier, and my comments were based upon our consolidated OpEx. My comments were that we're expecting OpEx to grow at 15% sequentially quarter-to-quarter over the course of the fiscal year. And then with respect to the mix, the mix is going to be entirely a function of the composition and magnitude of the bookings, which, as we've articulated in the past, we expect that in the foreseeable future to be relatively lumpy, where we could have a substantially higher QCaaS enterprise mix in any given quarter than we have systems bookings.

And then, as Alan mentioned earlier, each one of these deals has or could have unique revenue recognition elements to it. For instance, the percentage of completion on a systems installation. So the answer is expect that mix to be lumpy.

This concludes our question-and-answer session. I would like to turn the conference back over to Dr. Baratz for any closing remarks.

Thank you. So in closing, let me reiterate that D-Wave is different. We're pulling away from the quantum computing pack as demonstrated by our undeniable commercial traction and our remarkable technical leadership. D-Wave is the only dual platform quantum computing company capable of delivering both annealing and gate-model systems. The only company with quantum computers that have demonstrated quantum supremacy on a useful real-world problem.

The only company that has customer applications in production now, the only company with highly differentiated gate-model technology that delivers the remarkable speed of superconducting and the fidelity of ion trap or neutral atom approaches, a powerful combination that positions D-Wave to win in the error-corrected gate-model race. 2026 is the year of D-Wave Quantum, the year we emerge as a defining company in the quantum era.

Thank you all for joining us today, and we look forward to updating you on D-Wave's progress in the coming months.

This concludes today's conference call. You may disconnect your lines. Thank you for participating, and have a pleasant day.

Good day and welcome to the D-Wave Third Quarter 2025 Earnings Call. [Operator Instructions] Please note this event is being recorded. I would now like to turn the conference over to Mr. Kevin Hunt, Senior Director of Investor Relations. Please go ahead, sir.

Thank you, and good morning. With me today are Dr. Alan Baratz, our Chief Executive Officer; and John Markovich, our Chief Financial Officer. Before we begin, I would like to remind everyone that this call may contain forward-looking statements and should be considered in conjunction with cautionary statements contained in our earnings release and the company's most recent periodic SEC reports.

During today's call, management will provide certain information that will constitute non-GAAP financial and operational measures under SEC rules, such as non-GAAP gross profit, non-GAAP gross margin, adjusted EBITDA loss, adjusted net loss, adjusted net loss per share and bookings. Reconciliations to GAAP financial measures and certain additional information are also included in today's earnings release, which is available in the Investor Relations section of our company website at www.dwavequantum.com.

I will now hand over the call to Alan.

Good morning, everyone, and thank you for joining us today. I'm really pleased to share D-Wave's third quarter 2025 results, which reflect the ongoing momentum we've seen across all key business metrics, including revenue, gross profit, bookings and a healthy cash balance.

I will walk you through several of our recent business and technical highlights, starting with our continued commercial traction. The United Nations declared 2025 to be the international year of quantum science and technology, and it is evident that the world is watching the quantum industry in general and D-Wave specifically. Our esteemed leadership team has been invited to speak at events across the globe in order to address the growing interest in our Quantum solutions. From Tokyo to Berlin and Taiwan to Miami, businesses, research institutions and governments around the world are eager to learn more about D-Wave, our incredibly powerful yet energy-efficient technology and the impact it is starting to have for customers right now in solving complex computational problems that are outside the reach of classical computers.

Just weeks ago, D-Wave announced his participation as a founder of the Q-Alliance, an initiative designed to create a quantum hub in Italy that advances scientific discovery industrial transformation and digital sovereignty in the country. A core objective of the Q-Alliance is the development of a state-of-the-art quantum computing and research facility in Lombardi, Italy. In support of that effort, D-Wave announced a EUR 10 million contract for a D-Wave Advantage2 annealing quantum computer in the region, ensuring accessibility for Italy scientific community, academia, and industry. In partnership with the Italian government and the Q-Alliance, the agreement includes acquisition of 50% capacity of an Advantage2 system for 5 years with the option to purchase the full system. We expect to deploy the system sometime in 2026. While other foam companies are telling investors that sales really don't matter, we bank to differ.

Sales and customer success are key to business growth and driving shareholder value in the near and long term. Our market presence allows us to learn directly from customers and rapidly enhance our systems to address their needs. D-Wave offers the only quantum computers that have demonstrated advantage on a useful real-world problem and can support customer applications in production today. Our customer portfolio includes one of the world's largest airlines, one of the world's largest chemical companies, one of the world's leading mobile carriers and one of the world's largest payment companies. So as other quantum companies remain in R&D mode, we are laser-focused on a path to profitability built on customer value. We signed a number of new and renewing customer engagements in the third quarter for both commercial and research applications. These engagements include one of the largest U.S.-based international airlines, SkyWater, the nation's largest pure-play semiconductor foundry. Japan Tobacco's Pharmaceutical division, which is exploring new Quantum AI applications in drug discovery, Yapi Kredi, one is the leading banks in Turkey and Korean won computing, a company specializing in quantum computing R&D, quantum security solutions and AI infrastructure in Korea. We continue to work with a myriad of organizations on quantum computing applications across a diverse set of use cases. Most notably in the past quarter, we worked with BASF, one of the world's leading chemical companies completing a proof-of-concept project, they use a hybrid quantum application to optimize manufacturing workflows in a BASF liquid filling facility.

The hybrid quantum technology set a new benchmark for manufacturing efficiency, allowing a reduction of production scheduling time from 10 hours to just seconds. One of their key operational challenges involves scheduling liquids unloading underfilling lines for customer orders across several different products. While this may sound like a simple problem, it is, in fact, a very complex optimization problem that involves dozens of single chemical tanker trucks and hundreds of customer orders on fulfillment lines that require careful timing and coordination. The challenges often exceeded the capabilities of a classical only optimization approach, and our solution outperformed substantially with latency reduced by 14%, set at the time reduced by 9% and tank unloading durations reduced by up to 18%. During our last earnings call, I spoke to you about a hybrid proof of technology with North Wales police to optimize deployment of petrol vehicles. I'm pleased to report that the PLT was successfully completed in the third quarter with a hybrid quantum application that significantly outperformed classical resorts.

Our hybrid solution enabled North Wales Police to respond to over 90% of incidents within their target response time, reduced average incident response time by nearly 50% and reduced police vehicle coordination time from 4 months to 4 minutes, which significantly improves real-time adaptability. North Wales Police noted that the application could be scaled nationally. It is a valuable example of how our hybrid quantum computing solutions are beginning to show real-world potential across private and public sectors. Earlier this year, we announced the successful completion of a proof-of-concept with Japan Tobacco's Pharmaceutical division. They use D-Wave's quantum technology and AI to improve the drug discovery process. We are now taking that work a step further with a second proof of concept with Japan Tobacco for molecular discovery through Quantum AI. They are running a significant number of problems to help expedite the drug discovery process and they are receiving subsecond responses from the D-Wave Quantum computer, which is leading to improved performance versus classical computing alone for the large language model training. This work demonstrates that these hybrid LLN models produce more valid generated molecules compared to classical only methods. And we believe that no other quantum computer on the market today can produce such results. This work had a significant impact on speed to market for new drugs, which in turn could drive better patient outcomes.

Our results show that quantum annealing is the most effective method of quantum optimization. A recent IBM study showed a family of multi-objective optimization problems where gate model quantum optimization could compete with classical approaches. We threw this problem at our Advantage2 processor and found that it was 1,000x faster than all the classical and quantum approaches in the IBM study, in addition to finding higher quality solutions. You can read the full research paper on the archive. Against the backdrop of heightened global awareness around quantum computing and increasing exposure for D-Wave, we held our first-ever Qubits Japan User Conference in Tokyo in September. The response was fantastic. As Dr. Trevor Lanting, our Chief Development Officer, and I addressed a very enthusiastic audience representing many of Japan's leading companies and academic institutions. We were also honored to welcome Hidetoshi Nishimori as a presenter. Hidetoshi is widely recognized as the father of annealing quantum technology and the first to propose the concept nearly 30 years ago. Given the success of Qubits Japan, we are exploring hosting Qubits' events in other regions going forward, and our global Qubits 2026 conference will take place on January 27 and 28, 2026, and in just a couple of months in Boca Raton, Florida. Registration is now open and I hope to see many of you there in person.

Let me now turn to technology. As we drive important development work that is focused on helping customers realize the value of quantum computing now and in the future. Before I get to some specific product updates, I want to take a moment to discuss the underlying technologies or modalities that implement both of the 2 primary architectural approaches to quantum computing systems, that is annealing and gate model, as I believe there is a lot of misinformation and lack of understanding in this area. These technologies relate to how cubits are implemented. There are 4 primary methods for implementation, whether for gate or annealing. These implementation technologies are superconducting, high in track, neutral atom and photonics. We believe that one will clearly emerge victorious in the long run and that approach is superconducting. There are a few reasons why we believe superconducting will win. The first is gate speed with superconducting dates estimated to be 1,000 to 10,000x faster compared to the other major technologies. While approaches like ion trap or neutral atom may hold some near-term advantages in terms of Qubits Fidelity, we believe that gap will substantially close over time. We do not believe that the gate speed advantage will change materially.

So over the long term, superconducting is expected to have a massive performance advantage over competing approaches when one looks at the speed fidelity trade-off. We recently heard an intra company spent hours discussing their technology advantages at an analyst event, but not once did they mention gate speed. With a potential performance disadvantage of up to 10,000x, I can see why they might have forgotten to discuss that key metric. The second advantage for superconducting involves scalability and the role that manufacturing will play. Superconducting builds upon 50 years of integrated circuit manufacturing and packaging technology development that supports classical computing technology. Superconducting lever these established supply chains which we believe will provide the ability to scale faster and at a much lower cost. Other approaches require entirely new supply chains, which will likely require massive investments, extensive technological challenges and long time lines intel maturity as they attempt to scale. Those are the primary reasons why D-Wave chose superconducting technology for both our annealing and our gate model development programs. Unlike gate model systems, which will require error correction and thus are probably 5-plus years away from being truly commercial ready. Our annealing systems are available for commercial use today. It is also worth noting that much of the proven superconducting technology we develop for our annealing systems will likely also provide a competitive advantage for our gate model system. More than 60% of our patent portfolio relates to superconducting technology that we believe could apply in either a nailing or gate systems. The cryogenic control I discussed last quarter is a perfect example of a patented technology that to our knowledge, no other gate model company has today. and will almost certainly need to compete effectively with D-Wave's future gate model systems.

So now I'll turn to specific product development milestones. We're making solid progress in a number of areas, including our gate model program. As you are aware, D-Wave is the only company in the world building both a healing and gate model of quantum computers. So that's the only company that is currently positioned to address the entire market opportunity for Quantum. We see this as a competitive advantage that will give our customers quantum solutions that can address the full spectrum of their computational problems. As part of our gate model development initiative, we recently completed the fabrication of fluxonium qubit chips and superconducting control chips, and we are now bonding the to demonstrate scalable control of gate model qubits. This is a very important advantage -- sorry, this is a very important advantage that D-Wave has over any of our competitors as we believe this work will enable the first ever scalable gate bottle system with cryogenic control. And why is that important? Well, to provide any real computational utility with a superconducting gate model system, you need scale. And we believe cryogenic control provides the fastest path to large-scale gate model technology.

On the annealing quantum computing side of our business, earlier this week, we announced that the Advantage2 system installed at Davidson Technology headquarters in Huntsville, Alabama is now operational. This is a significant step in advancing the U.S. government's near-term use of D-Wave's quantum computing technology. The system is capable of addressing mission-critical computational problems that are beyond the reach of classical computers. Together with Davidson, we are already exploring quantum use cases in areas such as radar detection, resource deployment, military logistics optimization, material science and AI and look forward to continuing our collaborative work focused on national security and defense.

The Advantage2 system, which we made commercially available earlier this year is an engineering marvel, our most highly performing quantum computer yet and the only quantum computer that has demonstrated quantum supremacy on a useful real-world problem. A testament to its technical achievements, D-Wave was recently named as winner in Fast Company's 2025 Next Big Things in Tech Awards. This is a very prestigious award that recognizes emerging technologies with the potential to profoundly impact industries. D-Wave was acknowledged for showing what quantum computing can do right now. This is something we have been highlighting for several years given the production readiness of our technology compared to others and it is gratifying to see industry recognition that unlike all other quantum companies, D-Wave has commercial solutions capable of solving real-world problems today, not 5 or 10 years from now, but today.

Finally, turning to our annealing road map. Fabrication of our Advantage3 prototype chips is nearing completion, and we expect circuits for testing this quarter. As a reminder, our work on Advantage3 is focused on innovation and scaling, including increased connectivity and coherence, next-generation addressing and multi-chip processor fabric to accelerate our path to 100,000 qubits. In summary, the first 9 months of 2025 have been remarkable for D-Wave. We demonstrated quantum supremacy, sold our first Advantage system, introduced the Advantage2 system to market, further development of our gate model program, advanced the exploration of quantum and AI, worked with research and commercial customers on a variety of groundbreaking applications that go beyond the reach of classical, increased our cash position by over $650 million and much more. Our pipeline remains strong with large opportunities for both system sales and quantum compute as a service deals. And with more than $800 million in cash on our balance sheet, we remain well positioned to expand our business both organically and through M&A. We look forward to seeing many of you at upcoming investor conferences and in January in Boca Raton at Qubits 2026.

But before I hand it over to John, there's one more thing that I want to say. We recently had a fair amount of chest pounding from quantum leaders. Let me be clear. Anyone who characterizes quantum annealing as not real quantum is either intellectually incapable of understanding the physics and science or has chosen to put their head in the sand because they are worried about the competitive threat. Let's look at the facts. There is only one quantum computer in the world that has demonstrated the ability to solve an important useful problem that can't be solved classically, not a synthetic problem, but a useful problem, and that's our D-Wave Advantage2 system. When we announced this breakthrough work, there were research teams that tried to downplay the significance, but they never computed anything classically that we hadn't already computed classically and included in our science paper. Moreover, their scaling claims were ridiculous, and we have demonstrated that they are totally flawed through experimental results using their code, and this has been published on the archive. But it's not just that we are the only ones that have demonstrated true advantage. We are the only ones that have shown that we can do better than classical at all. Think about that.

There is no other quantum computer that has been able to demonstrate anything better than classical, let alone supremacy. For example, a quantum company that I mentioned previously recently touted multi-optimization results claiming as good as classical, not better, but as good as. We have run those same problems 1,000x faster than both their quantum computer and their classical approaches. You can find that on the archive as well. And it's not just about power. It's also about availability. Our systems are online with high uptime, providing subsecond response times. Other systems are frequently down. And when they are up, they regularly have multi-hour queuing delays. So let's ask this question, which systems are the real deal and which are toys and noisy toys at dat. Only D-Wave is the real deal.

With that, I'll hand the call over to John to provide a review of our third quarter and first 9 months of 2025 results. John?

Thank you, Alan, and thank you to everyone taking the time to participate in today's third quarter earnings call. In my review of the third quarter results, I will be providing non-GAAP operating metrics, including bookings, as well as non-GAAP financial metrics, including non-GAAP gross profit, non-GAAP gross margins adjusted net loss, adjusted net loss per share and adjusted EBITDA loss as we believe these metrics improve investors' ability to evaluate our underlying operating performance. These measures are defined in the tables at the bottom of today's third quarter earnings press release with the non-GAAP financial metrics for the most part, adjusting for noncash and nonrecurring expenses.

Revenue in the third quarter of fiscal 2025 totaled $3.7 million, an increase of approximately $1.8 million or 100% from the third quarter of fiscal 2024 revenue of $1.9 million. Third quarter revenue was comprised of $1.8 million in systems revenue associated with the upgrade of the Jülich Advantage System to an Advantage2 system. $1.4 million [indiscernible] revenue and $500,000 in professional services revenue. Bookings for the third quarter totaled $2.4 million, an increase of approximately $100,000 or 3% when compared to the third quarter of 2024 bookings of $2.3 million and an increase of $1.1 million or 80% compared to the immediately preceding fiscal 2025 second quarter bookings.

As Alan previously noted, after the close of the quarter, we signed a EUR 10 million agreement to place a D-Wave Advantage2 system in Europe, which will be reflected in our fourth quarter bookings that to date have totaled $12 million. This agreement is for 1/2 the capacity of an Advantage2 system over a 5-year period with an option to purchase the entire system at any time at a price that is within the range of our targeted system pricing of $20 million to $40 million. In terms of revenue recognition, the EUR 10 million will be recognized ratably over 5 years, commencing when the system becomes operational, which we expect to be sometime next year.

I would like to also reiterate my comments from the last 2 quarters' earnings calls, with respect to the composition of our sales pipeline incorporating incrementally larger average deal sizes than what we saw at this point last year. Two recent examples of this are the third quarter high 6-figure booking with a major U.S.-based international airline as well as the fourth quarter EUR 10 million European agreement booking. We continue to see increased activity from larger enterprises with more complex transaction structures. And as I have previously indicated, these deals typically take longer to close. That said, we remain confident in the outlook for booking activity going forward.

With respect to customers over the most recent 4 quarters, D-Wave had an excess of 100 revenue-generating customers, including approximately 2 dozen Forbes Global 2000 companies. GAAP gross profit for the third quarter of fiscal 2025 is $2.7 million, an increase of $1.7 million or 156% from the fiscal 2024 third quarter gross profit of $1 million, with the increase due primarily to the growth in revenue as well as higher margin systems upgrade revenue.

Non-GAAP gross profit for the third quarter of fiscal '25 was $2.9 million, an increase of $1.6 million or 131% from the fiscal 2024 third quarter non-GAAP gross profit of $1.3 million. GAAP gross margin for the third quarter of fiscal '25 was 71.4%, an increase of 15.6% from the fiscal 2024 third quarter GAAP gross margin of 55.8% in with the year-over-year improvement primarily driven by the growth in revenue and the higher-margin lit systems upgrade revenue. Non-GAAP gross margin for the third quarter of fiscal '25 was 77.7%, an increase of 10.5% from the fiscal 2024 third quarter non-GAAP gross margin of 67.2%. The difference between GAAP and non-GAAP gross profit and gross margin is limited to noncash stock-based compensation and depreciation and amortization expenses that are excluded from the non-GAAP measures.

Net loss for the third quarter was $140.8 million or $0.41 per share compared with a net loss of $22.7 million or $0.11 per share in the fiscal 2024 third quarter. The increase is due primarily to $121.9 million in noncash nonoperating charges related to the remeasurement of the company's warrant liability as well as realized losses stemming from warrant exercises, both of which increased materially due to the significant increase in the price of the company's common stock and warrants. Adjusted net loss for the third quarter was $18.1 million or $0.05 per share, a decrease of $5.1 million or $0.07 per share compared with an adjusted net loss of $23.2 million or $0.12 per share in the fiscal 2024 third quarter. The difference between net loss and adjusted net loss is primarily the noncash nonoperating warrant related charges. The adjusted EBITDA loss for the third quarter of fiscal '25 was $20.6 million, an increase of $6.8 million or 49% compared with adjusted EBITDA loss of $13.8 million in the fiscal 2024 third quarter. The increase was primarily due to higher operating expenses, partially offset by higher gross profit.

I'll now address D-Wave's operating performance for the first 9 months of fiscal 2025. Revenue from the 9 months ended September 30, 2025, was $21.8 million, an increase of $15.3 million or 235% from revenue of $6.5 million for the 9 months ended September 30, 2024. The year-over-year increase is largely due to the system sale to that has been recognized over the first 3 quarters of fiscal 2025. Year-to-date revenue was comprised of $15.5 million in revenue from the sale of an advantage annealing system to Jülich along with the associated advantage to processor upgrades. $42 million in [indiscernible] revenue and $2.1 million in professional services revenue. Bookings for the 9 months ended September 30, 2025, were $5.3 million, a decrease of approximately $300,000 or 7% from bookings of $5.6 million for the 9 months ended September 30, 2024.

GAAP gross profit for the 9 months ended September 30 was $18.5 million, an increase of $14.4 million or 353% from $4.1 million and GAAP gross profit for the 9 months ended September 30, 2024, with the increase due primarily to the recognition of the higher-margin system sale during the first 9 months of fiscal 2025. Non-GAAP gross profit for the 9 months ended September 30, 2025, was $19.2 million, an increase of $14.5 million or 304% from the non-GAAP gross profit of $4.7 million for the 9 months ended September 30, 2024.

Moving on to gross margins. GAAP gross margin for the 9 months ended September 30, 2025, was 84.8%, an increase of 22.1% from the 62.7% GAAP gross margin for the 9 months ended September 30, 2024, with the increase due primarily to the higher margin nature of the system sale. Non-GAAP gross margin for the 9 months ended September 30, 2025, was 87.8%, an increase of 15.1% from the 72.7% non-GAAP gross margin for the 9 months ended September 30, 2024. Net loss for the 9 months ended September 30, 2025, was $312.7 million or $1.01 per share compared with a net loss of $57.8 million or $0.32 per share in the fiscal 2024 9-month period. The increase is due primarily to $260 million in noncash nonoperating charges related to the remeasurement of the company's warrant liability as well as realize losses stemming from more exercises. Adjusted net loss for the 9 months ended September 30, 2025, was $52.8 million or $0.17 per share, a decrease of $5.1 million or 8.7% when compared to the fiscal 2024 9-month period net loss of $57.8 million or $0.32 per share. Adjusted EBITDA loss for the 9 months ended September 30, '25 was $46.7 million, an increase of $6.1 million or 15% from an adjusted EBITDA loss of $40.6 million for the 9 months ended September 30, 2024. The increase is due primarily to higher operating expenses, partially offset by higher gross profit.

Now a few comments on how D-Wave's revenue and business model are fundamentally different from most all, if not all, other quantum computing companies. Our revenue model now consists of 3 primary synergistic components. Quantum computing as a service or QCaaS professional services and system sales. Our quantum computing as a service consists of cloud-based recurring subscription revenue derived by providing customers with access to our own lead cloud service. While we have over 100 QCaaS customers, nearly 48% of which are commercial organizations. We believe that we are still at a very early stage in developing the QCaaS business with most of the revenue base still comprised of smaller deal sizes. Moreover, with 4 production systems now supporting the LEAP cloud service, we have over $100 million in annual QCaaS revenue capacity that inherently provides us with significant operating leverage.

Our professional services revenue typically involves relatively straightforward, fixed priced short-term engagements to assist customers with proof of concepts of applications that will help solve their business problems. Our objective is to transition these applications into longer-term production applications where customers access our systems on an ongoing basis to solve their computational problems. D-Wave is the only quantum company with applications in production. System sales is our newest revenue stream, and it is worth noting that our system sales are quite different from the systems being sold by most other quantum computing companies. Our advantaged systems are highly scalable, commercial-grade systems being used to solve real problems, whereas most other quantum computing systems our development stage low cubic count systems that are being used for research experimentation. While our systems may be used for research in areas like AI, these are not experimental or R&D systems.

Year-to-date, we have sold a system to the Jülich supercomputing center in Germany, entered into a memorandum of understanding the seller system to Jansa University in Incan, Metropolitan City and South Korea, signed a EUR 10 million contract for the D-Wave advantage to annealing and quantum computer to be located in Europe in partnership with the Italian government and the Q-Alliance. And earlier this week, we announced that an Advantage2 quantum computer is now operational at Davidson Technologies headquarters in Huntsville, Alabama to support U.S. Department of Defense and aerospace customers.

D-Wave's revenue model contrasts sharply with the revenue makeup of most other independent quantum computing companies typically comprised of highly concentrated government-funded research and development that is recognized under GAAP as revenue. In these arrangements, government entities are essentially funding the development of systems that they may or may not eventually purchase. This government-funded R&D revenue is typically low margin, and the systems involved are not commercial-grade, scalable production systems capable of solving real-world problems. They are experimental R&D systems. While we believe that increased government funding and Quantum could provide D-Wave with incremental revenue opportunities and could speed up development in certain areas. We don't believe that a primary focus on government-funded R&D revenue is a sustainable business model. We will continue to focus on providing access to our commercial-grade systems, either through our LEAP cloud service or via on-premises sales as well as providing the suite of professional services frequently necessary to transition applications into production.

Now I will move on to address the balance sheet and liquidity. As of September 30, 2025, D-Wave's consolidated cash balance totaled $836.2 million, representing a 2,700% increase from the year earlier fiscal 2024 third quarter consolidated cash balance of $29.3 million and a 2% increase from the fiscal 2025 second quarter consolidated cash balance of $819.3 million. During the quarter, the company received $40.3 million in proceeds from the exercise of warrants. On October 20, we announced the redemption of all the company's approximately 5 million outstanding warrants that if fully exercised before the redemption date would provide approximately an additional $58 million in cash. Subsequent to the end of the third quarter and through November 4, we have received $21.3 million from the exercise of warrants.

To conclude, our business momentum continues to build. And as we have previously stated, we believe that D-Wave has the opportunity to be the first independent publicly held quantum computing company to achieve sustained profitability and to achieve this milestone with substantially less funding than required by any other independent publicly held quantum computing company.

With that, operator, please open the call for questions.

[Operator Instructions] Our first question will come from Harsh Kumar with Piper Sandler.

Alan and John, first of all, congratulations on the Italy Lombardy deal and then also the very substantial 100 customer number that you mentioned on this call and the press release. I -- my question is, I see that you are now finally getting some attention/loved from the U.S. government. This has been something that has not happened in the past. You were getting attention from foreign governments, but not the U.S. government. I saw a deal where you're part of an alliance for national security work. Obviously, I won't ask about what you're doing, but can you talk about the significance here? And is this a shift towards some attention towards an easing by the U.S. government?

So Harsh, the key point that I want to make relative to the U.S. government is that our approach is quite different from the other quantum computing companies that are engaged with the U.S. government. We are not looking for R&D funding to aid in the development of our quantum computers. Our annealing quantum computers are at scale today and capable of solving important government problems. And so our work with the U.S. government is all about identifying areas where our systems can actually provide value to the government, whether it's in areas like military logistics or research placement or equipment maintenance. Those are the sorts of things we're focused on. And I think they're starting to become a realization in key areas of the government that our systems are capable of delivering value, and that's what's starting to open up some opportunity for us.

Next question will come from Troy Jensen with Cancer Fitzgerald.

Congrats on all the progress here on the green milestones. A quick question for Alan, and I love the passion here and all the speaking earlier, but can you just give us more details on the gate model product in the superconducting computer you're working on? It's specifically like the time line of the launch and kind of what the fidelity and cut count would be?

Yes. So first of all, we are using superconducting for our gate model quantum computer. We are not using the types of bids that most of the other superconducting gate model companies are pursuing. They are pursuing what are known as transmin qubits or voltage-controlled qubits, we are pursuing fluxonium qubits that are controlled with magnetic flux. We're doing this because our annealing quantum computers use flex-based qubits. And so we've got a lot of experience with them, and we think they have some very significant advantages over the transmin qubits when it comes to scaling. And then, of course, we are the only company that has been able to use cryogenic control with their client of computers I talked about this last quarter. That's why, for example, we can control 4,000 qubits in our annealing system with only 200 I/O lines rather than needing 3 to 5 aisle lines for qubit, and we're bringing that into the gate model space. our first demonstration will be what I discussed a bit earlier in the call, which is that we have now fabricated high-quality flexonium qubit chips, and we have fabricated cryogenic control chips, and we will be bonding those together through a bump on process to demonstrate scale -- basically cryogenic control of gate model qubits. From there, we will be moving on to developing basically logical qubit structures starting with small surface code but developed in a scalable fashion and then building up from there. And over the course of the coming year, we expect to have the first of this kind of capability demonstrable. And then that is on the path to scaling the logical cubics and the number of qubits on the chip to -- on the path to a scaled error-corrected gate model system. But the time line for a scaled [indiscernible] gate model system is still a number of years out. It's pretty much the same time line as you'll hear from any of other quantum computing companies who are being honest with you in the 5- to 10-year time frame.

The next question will come from David Williams with Benchmark.

I wanted to ask a little bit on -- and you talked about this earlier, but the Davidson relationship that you have. And I know you've been working on several projects there. But as we kind of think about national security and defense initiatives and things like the Golden Dome, how do you think you can play into that? Just kind of given your heritage there, and you've been working on this for a long time. It seems like you'd have a good opportunity. So just any color around that would be very helpful, I think.

Yes. So look, we are just as interested in Golden Dome as every other quantum computing company, we participate in the various government foreigns to kind of talk about this we're doing exploratory work in application areas that could be a benefit, but this really does fall right into the category of solving hard problems today that can deliver value to the U.S. government as opposed to trying to play with research systems, the new research experimentation. Our work with Davidson cuts across a number of different application areas, as I previously mentioned. And very interestingly, the next step in that relationship now that our system is fully calibrated operational and online at Davidson. The next step is to secure the system. So that we can run classified versions of applications on the system, which would make it what we believe would be the very first quantum computer certified for classified government applications.

Next question will come from Quinn Bolton with Needham & Company.

Congratulations on the momentum. I just wanted to follow up on the Q-Alliance transaction and SQD where they have the option to purchase the system. John or Alan, can you just sort of talk to the extent that they want to make that purchase, how would that work? You've got a EUR 10 million contract for half of the system over a 5-year period, would they get credit for money spent if they converted, say, in year 1 or year 2? Or would it be sort of a full $20 million to $40 million purchase price? And anything that had been recognized on the EUR 10 million contract you would keep and then you would get the full system sale to the extent that they move in that direction?

Yes. Sorry -- I'm sorry, Quinn, but we haven't closed any of the details around how the purchase would occur. The only thing I will say is that you should certainly not assume that $10 million for half the capacity of system means double that or $20 million for the full purchase of the system. There are many benefits that a customer gets when they purchase the system, and we pass title to them, including the ability to do things that with this system that can't be done when the system is online and shared. And so as a result, there's a premium on the pricing for the actual purchase of a system. But the relationship with the Q-Alliance and the Italy government is for 50% capacity of the system over the course of the next 5 years, and they do have the option to purchase the full system, which would then put them into a different category where title would actually pass.

Next question will come from Kingsley Crane with Canaccord Genuity.

Just to double-click on that Q-Alliance deal, it sounds like somewhat of an innovative deployment and procurement model. So I believe that 50% of the capacity would then be available to QCaaS customers. Could this become a blueprint to future deals, establishing points of presence across the world. Just curious your thoughts on the structure.

It absolutely could. And in fact, it's not entirely new for us. we did something like this with the Jülich Supercomputing Center. Initially, they had a system on site, and they had a portion of the capacity of that system, and the rest was available for other QCaaS customers, but then they chose to purchase the system. And so this is a really interesting model for a kind of building out presence around the world; and b, taking it one step at a time toward the purchase of the system. Of course, then there are other companies that just outright want to purchase the system, which is the nature of the discussion that we're having, for example, in Korea, and we have actually a few others of those that have now become quite advanced as well.

Your next question will come from Craig Ellis with B. Riley Securities.

Congratulations on progress with large customers and advantage to engagement. The question I wanted to ask was back to an expectation that was set 3 months ago, which was that the company would increase investment moderately about 15% half-on-half in R&D and sales and marketing as it looks to accelerate technology and its go-to-market capability. Can you just talk about how that's going? What we should expect in the fourth quarter with respect to operational things, increased capability? And what should we expect as we look at the first half of '26, more stable expenses or a further uptick?

John, do you want to talk about the numbers and then I can maybe provide a little color on the use of the funds?

Sure. So as we outlined in the second quarter earnings call, we outlined that our operating expenses for the balance of the year will increase approximately 15% on a sequential basis with the majority of that incremental spend in the R&D area. Craig, with respect to your question in the first half of next year, we're not going to provide any guidance beyond what we have provided for the balance of this year in terms of the growth in the OpEx.

And Craig, just to get a little color on you. So I think that it won't be a surprise to anybody if I say that given the fact that we have significantly more cash now than in the past. And so we do have the ability to start making some modest investments in R&D. One of the key areas that we will be accelerating is our gate model program. We believe that we've got some unique, valuable technology in the gate space that's going to allow us to move in a fairly short footed fashion. And so that is a target for a bit of increased investment. And then from a go-to-market perspective, now that we have the system sales model, we are -- we've got a portion of the team that's now focused on system sales and a portion that's focused on professional services in QCaaS. So we've made a bit of an investment to kind of facilitate both of those go-to-market motions. And then we're starting to see something that is actually pretty exciting. It's still early with respect to exactly when and how this will materialize. But we've talked a lot about proof of concept and the value that we're seeing and the move to production. And we've started to have the first conversations along the lines of, "Well, hey, this application is looking really good." We want to move it into production, but we've got a few other applications now that we're interested in. Could we do a deal that gives us the ability to support multiple proofs of concept and multiple applications in production. I'm not entirely sure what to call it, but you could think of it as maybe something like an enterprise license that bundles a number of production apps and a number of proofs of contact to really start accelerating the professional services and QCaaS sales and revenue, that's just getting started. I mean -- I mean that's happened literally within the last few months where we started getting inquiries about that. But that could be an important inflection point for us on the PS QCaaS business.

Your next question will come from Suji Desilva with ROTH Capital.

Congrats on the progress here. I wanted to ask a question about in the quantum market. There's a lot of discussion of ancillary quantum opportunities. communications, timing, memory, security keys and so forth. I just want to understand qubits position on these? Are you razor-focused on your core opportunity? Are you able to internally develop these inorganic or maybe the markets are not as mature because I think just any thoughts there would be helpful?

We are laser-focused on quantum computing. First of all, our systems for the foreseeable future don't need the quantum interconnect you hear about on quantum networking the other modalities, the other approaches do. For example, when it comes to interconnecting tracks, whether they be [indiscernible] or [indiscernible] track, you need that kind of interconnect and you're going to need to interconnect the traps to scale. But for our systems, that's not that's a critical technology. And we don't view sensing a central to what we're doing. We don't view quantum computing distribution essential to what we're doing. We are focused on quantum computing.

Your next question will come from Richard Shannon with Craig-Hallum.

This is Tyler on for Richard. So I wanted to double-click on some of the things that your customers are doing. You mentioned SkyWater is using your QPU, is this for magnetic material simulation or for the simulation of quantum devices? And does this help improve the ships that you're getting from them? And then also for the airline company, are they using your technology for scheduling or traveling salesmen? And then if you can have a comment on how often BASF has to run your program, if that is in production, I'd be interested in that.

Okay. I'm not sure whether to call that 1 question or 3 questions. So here's what I will say. In the case of SkyWater, they're a new customer as of this quarter, and it is really all about looking into kind of optimization problems in their operations. And in the case of this very large airline, we have not disclosed the applications that we are working on.

The next question will come from John McPeak with Rosenblat Securities.

It's good to hear things are tracking well. I just have a question on optimization. There's a lot of publicity around quantum, a lot of feverish commentary that you were referring to. And I'm curious relative to trial to conversion to production on the optimization side, new logos. Maybe you could just talk a little bit about what the tone is like there because that's real business.

Yes. I'm not into sure what you mean by the tone, but what I can tell you -- and we started to see this maybe a quarter or 2 quarters ago, we are now being engaged by much larger companies interested in addressing larger and more challenging optimization problems. And that's why we can talk about one of the largest airlines, one of the largest chemical companies, one of the largest pain processing companies, that's why I commented that we're now with successful initial proof of concept starting to get inquiries about, okay, if I wanted to do a deal that bundles in multiple proofs of concept and assuming they're a successful multiple production applications. Could I -- what would a deal like that look like. So I think, as I said before, we're getting close to an inflection point on the professional services and QCaaS business, where we're seeing now better and better results with our customers, due in large part to the fact that we're now on the Advantage2 platform. We now have our really powerful NL hybrid solver. And so the results are, in terms of speaking for themselves and allowing us to move forward more aggressively.

This concludes our question-and-answer session. I would like to turn the conference back over to Dr. Alan Baratz for any closing remarks. Please go ahead.

Okay. Thank you, operator. Momenta is clearly building at D-Wave, and we believe that our first-mover advantage is increasingly evident in our business results and our technical innovation. As Fox businesses Charles Payne remarked earlier this week, we are the real deal. In the seal quantum pipe, our position is clear. We're helping customers realize the value of quantum today, and I think our 2025 results to date demonstrate that we are making consistent progress toward the service of that mission. So thanks for your time today, and thanks for your support, and I look forward to updating you again in January at qubits 2026. We'll see you all on both of your time. Thank you.

The conference has now concluded. Thank you for attending today's presentation. You now disconnect.

Good morning, everyone, and welcome to D-Wave's Second Quarter of Fiscal Year 2025 Earnings Conference Call. Today's conference call is being recorded.

At this time, I would like to turn the call over to [ Beth Mathis ] of Investor Relations. Please go ahead.

Thank you, and good morning. With me today are Dr. Alan Baratz, our Chief Executive Officer; and John Markovich, our Chief Financial Officer.

Before we begin, I would like to remind everyone that this call may contain forward-looking statements and should be considered in conjunction with cautionary statements contained in our earnings release and the company's most recent periodic SEC reports.

During today's call, management will provide certain information that will constitute non-GAAP financial measures under SEC rules, such as non-GAAP gross profit, non-GAAP gross margin and adjusted EBITDA loss and operating metrics such as bookings. Reconciliations to GAAP financial measures and certain additional information are also included in today's earnings release, which is available on the Investor Relations section of our company's website at www.dwavequantum.com.

I will now hand the call over to Alan.

Good morning, everyone, and thank you for joining us today. I'm once again really excited to share our results for the second quarter of fiscal 2025. Building on the company's remarkable first quarter results, we continue to see accelerating momentum across the business.

Let me now walk you through some key highlights, starting with technical achievements. In May 2025, we announced the general availability of D-Wave's Advantage2 quantum computer, our most advanced and performance system. The Advantage2 system is a powerful and energy-efficient annealing quantum computer capable of solving computationally complex problems beyond the reach of classical computers. A smaller prototype of this system was used to demonstrate our quantum supremacy result on a real-world materials simulation problem, a first for the industry.

Featuring D-Wave's most advanced quantum processor to date, the Advantage2 system is commercial grade and built to address real-world use cases in areas such as optimization, material simulation and artificial intelligence.

As previously shared, the Advantage2 quantum processors have demonstrated impressive performance gains over the previous Advantage system, including double coherence time for faster time to solution, a 40% increase in energy scale for higher quality solutions and increased qubit connectivity from 15 to 20 way to enable solutions to larger problems.

It's a significant engineering achievement that highlights our progress in scaling quantum technology to meet demands for growing computational processing power while maintaining energy efficiency. We're helping customers realize value from quantum computing right now, and the Advantage2 system is an important proof point.

We recently announced a new strategic development initiative focused on advanced cryogenic packaging, designed to advance and scale both gate model and annealing quantum processor development. The initiative builds on D-Wave's technology leadership in cryogenic packaging and will expand our multichip packaging capabilities, equipment and processes. By bolstering D-Wave's manufacturing efforts with state-of-the-art technology, the company aims to accelerate its development efforts in support of its aggressive product road map on the path to 100,000 qubits.

As part of this initiative, D-Wave is leveraging deep expertise and processes at the NASA Jet Propulsion Laboratory, JPL. Harnessing JPL's superconducting bump on process, we have demonstrated end-to-end superconducting interconnect between chips, work that we expect will serve as an important foundation for scaling both our annealing and our gate model systems.

We are continuing important development work that brings together quantum and AI to explore the synergies and benefits of these complementary technologies. Our aim is to help organizations accelerate the use of annealing quantum computers in a growing set of AI applications. To that end, we recently introduced a collection of developer tools to advance quantum AI and machine learning innovation.

First, we launched an open source quantum AI toolkit that provides direct integration between D-Wave's quantum computers and PyTorch, a production-grade machine learning framework widely used to build and train deep learning models.

Second, we launched a new demonstration that illustrates how developers can use D-Wave's quantum AI toolkit to generate simple images, reflecting what we believe is a pivotal step in the development of quantum AI capabilities.

These tools are making it easier for customers like Japan Tobacco and Triumph to build hybrid quantum classical machine learning applications. Customers are increasingly coming to us to explore how to integrate quantum into AI workflows, and we expect that this will remain a priority development area for us.

Last quarter, we discussed the purchase of an Advantage2 system by the Julich Supercomputing Center, an important milestone in our burgeoning on-premises business. Demand for purchasing a system has been high, and we've been in discussions with numerous organizations around the world interested in buying a D-Wave quantum computer.

Recently, we announced a strategic relationship with Yonsei University and Incheon Metropolitan City to accelerate the exploration, adoption and usage of quantum computing in South Korea. Under the terms of the memorandum of understanding, the 3 organizations are working together to advance research and talent development for quantum computing to provide access to D-Wave's quantum computing systems and services and to collaborate on the development of use cases in biotechnology, material science and other areas. In addition, the MOU supports our efforts toward the acquisition of a D-Wave Advantage2 system on site at the Yonsei University International Campus in Songdo, Yeonsu-gu, Incheon.

Before turning to commercial updates, I wanted to take a moment to remind everyone of the differences between annealing and gate model, as there still appears to be continued misinformation that we believe is confusing the market.

Annealing and gate model are different types of quantum computing approaches that solve different types of problems. According to peer-reviewed research, gate model quantum computers will not offer advantages for all problem types. Multiple research results have shown that annealing quantum computers outperform and are expected to continue to outperform gate model on optimization problems. So gate model cannot universally solve all problems better than classical.

It is also important to understand that annealing is not a niche solution. We believe that annealing quantum computing is well suited to a broad set of problems, including AI machine learning, quantum simulation and business optimization, which is ubiquitous in today's modern enterprise. Business optimization problems encompass things like workforce scheduling, production scheduling, resource allocation, vehicle routing and so on.

Together, these represent extensive use cases with far-reaching potential. To characterize annealing quantum computing as niche is misleading and ill-informed. Both annealing and gate model quantum computers can solve a broad range of problems and each has its limitations with problems it cannot efficiently solve.

Annealing quantum computing is very good at solving optimization problems, which cannot be efficiently solved by gate model. Gate model quantum computers, once commercialized, are expected to be very good at quantum chemistry and 3D fluid dynamics, which cannot be efficiently solved by annealing. And both systems, both systems can tackle linear algebra and factorization, meaning problems related to AI, machine learning and cryptography. We believe that organizations will need both annealing and gate model systems in order to address their full problem sets. This is why D-Wave is building both types of quantum computers for our customers.

Now I'll turn to commercial updates. In terms of D-Wave's customer portfolio, we signed a number of new and renewing customer engagements for both commercial and research applications, including E.ON, a European multinational electric utility company; GE Vernova, a global energy company; the National Quantum Computing Center, NTCC, the U.K.'s National Lab for Quantum Computing; Nikon Corporation, a multinational corporation specializing in optics and precision technologies; NTT Data Corp., a multinational IT services and consulting company; NTT DOCOMO, Japan's leading mobile operator; Sharp Corporation, a multinational electronics company; and the University of Oxford.

We've also been working with a Fortune 500 aerospace and defense company. And in Q2, we completed a prioritization of 12 different use cases applicable to their business operations that the customer found challenging to solve using classical optimization techniques. Quantum optimization, powered by annealing quantum computing can deliver value in terms of better and faster solutions. Based on the results of our initial exploratory work with this customer, we have now started building the proof of concept for the first of the use cases with a road map to expand to all of them and market them to additional aerospace companies.

In addition, we recently built a quantum hybrid proof of technology with North Wales Police to optimize the deployment of patrol vehicles. The proof of technology solution was tested against historical data and exceeded the customers' expectations, meeting target response times for more than 90% of incidents and using just 10 seconds of solve time. We're encouraged by these initial results and see them as important proof points of quantum hybrid's potential for law enforcement-related use cases.

We're also seeing growing interest in the Leap Quantum LaunchPad program, which is a 3-month trial that provides access to D-Wave's quantum computing systems, our Leap real-time Quantum Cloud service and our team of quantum experts for project support. Since its introduction in January of 2025, the LaunchPad program has received more than 1,300 applications spanning business, government and academic institutions. The program is serving as an important vehicle to attract and fast-track customers into proof-of-concept development and ultimately, application deployments.

So to summarize, we are continuing to make steady progress across our business. First, delivering on technical milestones, including the release of our sixth-generation quantum computing and advancing quantum AI development.

Second, executing against our go-to-market strategy, including increased discussions for on-premises systems with a variety of interested parties.

And third, working closely with customers to develop hybrid quantum applications that are addressing critical organizational problems.

With the strongest cash position in our company's history, we believe that we are very well positioned to explore M&A activity that will propel our business even further and faster while delivering value to customers and shareholders alike.

With that, I'll hand it over to John to provide a review of our second quarter fiscal year 2025 results. John?

Thank you, Alan, and thank you to everyone taking the time to participate in today's call. In my review of the second quarter and first half results, I will be providing non-GAAP operating metrics, including bookings as well as non-GAAP financial metrics, including non-GAAP gross profit, non-GAAP gross margins and adjusted EBIT loss as we believe these measures improve investors' ability to evaluate our underlying operating performance. These measures are defined in the tables at the bottom of today's second quarter earnings press release with the non-GAAP financial measures, for the most part, adjusting for noncash and nonrecurring expenses.

Revenue for the second quarter of fiscal 2025 totaled $3.1 million, an increase of about $900,000 or 42% from the second quarter of fiscal 2024 revenue of $2.2 million. The second quarter revenue includes $1 million in revenue associated with the Advantage2 quantum processing unit upgrade for the annealing system that was installed at the Julich Supercomputing Center in the first quarter of this year.

Revenue from the Advantage2 upgrade is recognized using the percentage of completion method, reflecting the timing of installation services that are closely integrated with the QPU or quantum processing unit hardware. We expect that this upgrade will be substantially complete by the end of this year.

Bookings for the second quarter totaled $1.3 million, an increase of approximately $600,000 or 92% from the second quarter of 2024 bookings of $700,000. As we have previously mentioned, we are encouraged by an expanding sales pipeline that includes a growing number of large enterprises and well-known logos with a market increase in the size of the average transaction size when compared to a year ago. Many of these companies are focused on having us build them proof of concepts versus just buying a small amount of QCaaS or quantum compute-as-a-service that translates to incrementally more complex transaction structures.

This, in combination with the challenges associated with dealing with substantially larger organizations with multistep and sometimes very rigid procurement processes and documentation requirements has resulted in deals taking longer to close than what we had originally anticipated. Over the last 4 quarters, we had over 100 revenue-generating customers that includes customers in the commercial, government and research sectors and nearly 2 dozen Forbes Global 2000 companies.

GAAP gross profit for the second quarter was $2 million, an increase of approximately $600,000 or 42% from the second quarter of fiscal 2024 GAAP gross profit of approximately $1.4 million, with the increase due primarily to the growth in revenue.

Non-GAAP gross profit for the second quarter was $2.2 million, an increase of approximately $600,000 or 39% from the second quarter of fiscal 2024 non-GAAP gross profit of approximately $1.6 million. The difference between GAAP and non-GAAP gross profit and gross margin is limited to noncash stock-based compensation and depreciation expenses that are excluded from the non-GAAP gross profit and gross margin.

GAAP gross margin for the second quarter was 63.8%, representing a slight improvement from the second quarter of fiscal 2024 GAAP gross margin of 63.6%. Non-GAAP gross margin for the second quarter was 71.8%, a slight decrease of 1.3% from the second quarter of fiscal 2024 non-GAAP gross margin of 73.1%.

Net loss for the second quarter was $167.3 million or $0.55 per share, an increase of $149.5 million or $0.45 per share from the second quarter of fiscal 2024 net loss of $17.8 million or $0.10 a share. The increase in the net loss was primarily due to $142 million in noncash nonoperating charges related to the remeasurement of the company's warrant liability as well as realized losses stemming from actual warrant exercises.

In extracting the impact of the noncash nonoperating warrant remeasurement and related charges from the GAAP net loss, the adjusted net loss for the second quarter was $25.3 million or $0.08 per share, an increase of $5.3 million and a decrease of $0.04 per share from the fiscal 2024 second quarter adjusted net loss of $20 million or $0.12 per share.

Adjusted EBITDA loss for the second quarter was $20 million, an increase of $6.1 million or 44% from the second quarter of fiscal 2024 adjusted EBITDA loss of $13.9 million, with the increase due primarily to higher operating expenses that is reflective of our investments to support our future growth opportunity, partly offset by higher gross profit.

I'll now address the performance for the first half of the year. D-Wave's revenue for the 6 months ended June 30 was $18.1 million, an increase of $13.5 million or 289% from revenue of $4.6 million in the 6 months ended June 30, 2024.

Bookings for the first half of fiscal 2025 were $2.9 million, a decrease of approximately $400,000 or 13% from bookings of $3.3 million in the first half of fiscal 2024.

GAAP gross profit for the first 6 months of fiscal '25 was $15.9 million, an increase of $12.9 million or 420% from GAAP gross profit of $3 million for the first 6 months of fiscal 2024, with the increase due primarily to the high-margin system sale during the 6 months ended in June.

Non-GAAP gross profit for the first 6 months of fiscal ' 25 was $16.3 million, an increase of $12.8 million or 367% from the year earlier 6 months non-GAAP gross profit of $3.5 million.

GAAP gross margin for the first half of '25 was 87.6%, an increase of 22% from the 65.6% GAAP gross margin in the first half of fiscal 2024, with the increase due, again, primarily to the high-margin system sale during the first 6 months -- for the 6 months ended in June. Non-GAAP gross margin for the first half of fiscal '25 was 89.9%, an increase of 14.9% from 75% in the 6 months ended June 30, 2024.

Net loss for the 6 months ended June 30, 2025, was $172.8 million or $0.59 per share compared with a net loss of $35.1 million or $0.21 per share for the 6 months ended June 30, '24. In adjusting the impact of the noncash nonoperating warrant remeasurement and related charges from the GAAP net loss, the adjusted net loss for the 6 months ended June 30 was $34.6 million or $0.12 per share, essentially flat when compared to the adjusted net loss of $34.6 million or $0.21 per share for the 6 months ended June 30, 2024.

Adjusted EBITDA loss for the first half of fiscal '25 was $26.1 million, a decrease of approximately $700,000 or 3% from the adjusted EBITDA loss of $26.8 million in the first half of 2024, with the improvement due primarily to higher gross profit, partially offset by increased operating expenses.

Moving on to the balance sheet and liquidity. As of June 30, D-Wave's consolidated cash position totaled a record $819.3 million, representing over a 1,900% increase from the fiscal 2024 second quarter consolidated cash balance of $40.9 million and a nearly 170% increase from the immediately prior fiscal 2025 first quarter consolidated cash balance of $304.3 million.

During the second quarter of fiscal 2025, we raised over $500 million in equity, including $400 million in gross proceeds from our fourth at-the-market equity program, $99.3 million in net proceeds from the exercise of warrants and $37.8 million in net proceeds from our equity line of credit with Lincoln Park Capital Fund that fulfilled the $150 million commitment that was originally secured in June of 2022. Subsequent to the end of the quarter, we received an additional $15 million from the exercise of warrants.

Lastly, during the quarter, we fully recovered the $1 million investment plus accrued interest that we made in Zapata AI in February of 2024 through a convertible note instrument that we wrote off later that year when Zapata became insolvent.

As a result of the magnitude of capital that we have recently raised, in addition to pursuing strategic acquisitions, we are accelerating a number of our key investment initiatives. In the area of research and development, we are investing in superconducting bump bond process as highlighted in Wednesday's -- last Wednesday's press release and as Alan mentioned earlier. This process will support our multichip processor program on our path towards a 100,000 qubit annealing system. This process will also support scalable cryogenic control of fluxonium-based gate model technology.

We will also be upgrading our superconducting printed circuit board advanced packaging manufacturing operation and increasing the number and frequency of our wafer fabrication runs to support building Advantage3 prototypes as well as continuous improvements to qubit coherence times for both our annealing and gate model architectures. In addition, we will be investing in a number of quantum AI research and development programs.

In the area of sales and marketing, we will be expanding the size and geographical footprint of our professional services organization to support growing demand for quantum optimization customer engagements across both commercial and government sectors, including U.S. Defense.

And lastly, in the area of G&A, we will be making further investments in our cybersecurity personnel and infrastructure. For the balance of this year, we expect these incremental investments will result in a quarterly non-GAAP OpEx that is approximately 15% higher than our second quarter actual non-GAAP OpEx.

To conclude, as we have previously stated, we believe that D-Wave has the opportunity to be the first independent publicly held quantum computing company to achieve sustained profitability and achieve this milestone with substantially less funding required by any other independent publicly held quantum computing company.

Given that we are now fortunate enough to have 10 security analysts covering D-Wave, we will, in the essence of time, ask each analyst to commence the Q&A session with one question, and then we will go back through the queue for additional follow-on questions.

With that, we will now open up the call for questions.

[Operator Instructions] Our first question comes from Craig Ellis of B. Riley Securities.

Congratulations on the results in the quarter and technical progress, guys. I wanted to start by following up on the point the company had made about M&A and understand the types of M&A that the company believes would make sense, either items that are more technical in nature, maybe helping to accelerate the gate model side of the business or things that might be more go-to-market oriented? And then what size of M&A makes sense for the company? And finally, related to that, what are we looking at in terms of timing as we pursue that growth angle?

Craig, thanks for the question. So we have not disclosed our strategy and plans for M&A other than to say that with $800 million in the bank, it has now become a strategic priority for the company. That having been said, over the course of the last several months, we have been spending a fair amount of time developing a strategy and a plan. And it falls into a number of key areas, including some of those that you mentioned. But think about it as really accelerating our R&D and product development activities in a number of key areas, which could be everything from gate model to quantum AI.

And anything on timing there, Alan, whether we're looking at something that could be 2025 versus 2026?

It's hard for me to predict. But what I would say is that our goal would be to start being able to announce acquisitions this year. However, it takes 2 to tango. And so we'll just have to see how that plays out.

Our next question comes from David Williams of Benchmark.

Congrats on the continued progress here. So maybe if you could speak to the cryogenic news and the importance of that towards your road map and what you think it will bring ultimately as you get that ramped into the road map?

Yes, David, thank you. This is really important to us. So we've talked in the past about how we have significant intellectual property and a real lead in cryogenic control. And by that, I mean the ability to control qubits and control our systems on chip rather than needing to do all the control from room temperature. And now as we are looking to leverage that into the gate model program as well as expanding our annealing processes to much larger numbers of qubits, as we said, 100,000 qubits, we really need to be moving to a multichip solution.

And when we start interconnecting chips together in the refrigerator, we need to make sure that, that interconnect is also superconducting and that we can preserve the quantum properties like entanglement across those interconnects. And so this is key to both scaling our annealing systems as well as developing our gate model systems.

And frankly, we made progress in this area much faster than we actually expected it to. The NASA Jet Propulsion Laboratory had some capabilities that we thought looked interesting. We thought it would take a while to get to the point where we could actually kind of evolve that into what we needed. But it actually moved a lot faster than we thought. And so now we're really starting to build a production capability around that technology to more rapidly drive the multichip annealing processes as well as the gate model system.

Our next question comes from Troy Jensen of Cantor Fitzgerald.

Gentlemen, I also want to offer my congrats on all the technical milestones here. Maybe for you, Alan, I'd just love to hear more about Advantage2. And I guess a few things kind of all coupled together would be, is South Korea deployment, is that going to be Advantage2 or a number of systems you expect to install by the end of next year and maybe some of the technical improvements in the platform?

Okay, Troy. So first of all, Advantage2 is a really important system for us in the sense that it was the first system on which we were able to demonstrate true quantum supremacy, specifically the ability to solve a useful real-world problem on a quantum computer that cannot be solved classically, full stop. It is what everybody in the quantum industry aspires to, and we were the first to achieve it, and we achieved it on the Advantage2 system. I will tell you, we tried to get this result on our earlier Advantage system, and we were not able to. It took the increased capability of Advantage2 to be able to perform that computation. And specifically, it required the additional interconnect to more efficiently map the problem into the quantum processor. And it also took the longer coherence time and the increased energy scale to get the solutions faster and more accurately. So this is a significant advance over Advantage, and we're really excited about it. And it's also, I think, driving that increased customer interest that John talked about relative to much larger companies with much larger opportunities that are now engaged with us in a kind of sales cycle process.

As far as the number of systems, this falls into 2 categories. One is our Quantum Cloud service. And we have 4 production systems in our Quantum Cloud service today. We don't really need more than that in the cloud service for relatively near-term revenue growth. We likely will add a couple more systems down the road, but 4 is sufficient for now. And obviously, all 4 of those will be upgraded to Advantage2. Currently, one of them has been upgraded. Ultimately, all of them will be upgraded.

And then there's the on-premise systems. And obviously, Julich is one of those, and we are already in the process of doing that upgrade, as John pointed out. And then as we sell more on-premises systems, those will be Advantage2 processors. And we said that we've got a really good pipeline for sales of systems. We've got a second one that we're closing in on, which is the South Korea. We've got another one that's now starting to kind of emerge as a relatively near-term opportunity and then a pipeline of others. But in the past, I have said in the near term, think more like 1 a year than multiple a year. I'd still say that, although starting to feel like maybe it could be a little bit more. But the number is still kind of relatively small in the near-term. So that having been said, if you add up everything that I said, we're talking maybe 6 or 7 Advantage2 systems.

Our next question comes from Richard Shannon of Craig-Hallum.

This is Tyler Anderson on for Richard Shannon. So could you elaborate on the developer tools that you released? And noting that the problem that you demoed is a classical ML problem, can users leverage higher pixel images for this? And then also for the sake of timing question, you mentioned control right after your bump bonding. Do you have already or plan to have integrated control onto your chiplet such that there isn't an external control mechanism?

So Tyler, I will answer your first question to be fair to everybody else because we did say only one question, and then you can go ahead and get back into the queue.

With respect to what we announced as far as the AI developer toolkit, which I believe is what you're referring to, essentially, what we have done is we have introduced the ability to use PyTorch, which is an open source Python-based machine learning platform that's frankly in fairly wide use for training large language models. And we've introduced into that the ability to use a technology that we actually developed within D-Wave, which is called a Discrete Variational Autoencoder.

Now I'm not going to get into the technical details of what I mean by that other than to say what this does is allows you to take a data set and map it into a latent space, which is really what machine learning is all about, essentially map it into a machine learning model that can then be used to recreate the data and other things that look like that data set. But what's unique about this is that we're mapping it into a discrete latent space, not a continuous latent space. Machine learning today typically operates on continuous data, but we're now mapping it into a discrete latency, seeing excellent results with the discrete latent space. But what's so important about that is that our quantum processors natively work with discrete data, not with continuous data.

So what this does is it actually opens up the opportunity for the annealing quantum computer to be the vehicle for creating that discrete latent space or doing the learning, which we ultimately think could deliver better models faster and with lower energy consumption. But that having been said, I also want to point out that this is just the next step on our journey in the area of quantum AI. And there are a number of other things we are working on today in the lab that take this significantly further than what we've announced so far that we're very excited about as well.

Our next question comes from Suji Desilva of ROTH Capital.

I appreciate all the color thus far on quantum AI. Can you talk, Alan, about the next milestones or activities to watch indicating D-Wave's progress here?

In the area of quantum AI?

Correct.

I think the short answer to the question is no because we haven't yet announced anything beyond what we announced a few days ago. But just to kind of, say, a little bit -- follow-up a bit more on what I said a minute ago, there are a number of modern approaches to AI and machine learning model training and inference. Variational autoencoders is one approach, but there are other important things like transformation model -- sorry, transformer models and diffusion models. And we're working with those as well. And we expect that over time, we would be able to deliver a platform that could leverage our quantum systems in support of all of those approaches.

Our next question comes from Harsh Kumar of Piper Sandler.

Congratulations on a lot of progress, Advantage2 and other things that you're doing. Alan, I wanted to ask you about the toolkit information you provided in your press release. Is it possible that you could have libraries kind of like how NVIDIA does offer to its customers sort of hot finish models that they can -- the customers can then take and sort of finish up and customize. Is that sort of the idea? Or is that even possible with the toolkit with demos reference that you mentioned in your press release and commentary?

So it's absolutely possible. And in fact, the demo is one simple example of that. However, currently, we are working with customers, and leveraging their data in the application areas that are important to them. For example, Japan Tobacco and Triumph, the 2 that I mentioned a bit earlier, rather than trying to build these out ourselves. The extent to which, a, we take any of that and pull it back into our platform will depend a little bit on the customers and the extent to which we've negotiated the ability to be able to do that. And then whether we choose to start pursuing any applications or application templates ourselves is not something that we've certainly not announced it. We've not talked about it. I think that would require us to bring domain expertise in key problem areas into the company. It's something we're thinking about. But honestly, at this point, I wouldn't say we're going to do it. I would just say we're thinking about it.

Our next question comes from Kingsley Crane of Canaccord Genuity.

So quantum annealing really has significant potential with both corporates and nation, states or agencies. I'm curious how the tenor in the U.S. government has shifted, specifically with respect to quantum annealing in the past couple of quarters? And then just any thoughts on DARPA's Quantum Benchmarking Initiative and if there's an opportunity for annealing within that framework?

You really know how to push my buttons. Okay. Let's start with QBI. The DARPA QBI program is totally focused on gate model. And I think that is a huge mistake on the part of DARPA and the U.S. government. I think that by focusing on gate model, they are totally missing the fact that annealing is the most capable approach to quantum for many of the important problems that the government needs to address, whether it's in the area of defense for things like missile defense or troop resupply or in the area of transportation, for example, things like port logistics or we've done work in the area of wildfire fighting. I mean, the truth of the matter is annealing represents, I think, not only the best, but the only quantum approach that can address many of the government's hard computational problems. This is back to the fact that many of these are optimization problems, which require annealing quantum. Gate model cannot address them.

And by excluding annealing from the QBI program, I think DARPA has made a huge mistake. And I would encourage them to maybe not add annealing to the QBI program, but maybe create a second quantum program for non-gate model approaches to ensure that the U.S. government is kind of focused on all the approaches to quantum computing, not just one approach to quantum computing.

Now with respect to progress more broadly in the U.S. government, I'd say the answer is yes, but slow. So we are making inroads into different application areas because we uniquely can do that since we have a system that is capable of delivering value today, not 5 or 10 years from now, but it's slow going.

Our next question is coming from Ruben Roy of Stifel.

Alan, I don't mean to push your buttons here, but I'm going to ask this question anyway. So -- and I guess it's in the context of the M&A commentary that you made and also some of the comments that John made with respect to having conversations with larger customers and sort of getting feedback, I guess, from them. So with all of that in mind, I guess the simple question is, has your philosophy on the timing of when D-Wave might think about bringing a gate model to market changed? Has that accelerated for any reason or no? And if no, kind of are customer conversations sort of driving you to think that the time frame that you guys were thinking about previously is probably the right time frame?

Yes. That's not pushing my buttons at all. That's just a great -- well, they're all good questions and they're all good conversations to have, but I'm not annoyed about gate model and annoyed about DARPA. So relative to our gate model program, look, we still believe that, a, you will never, never see a commercially viable gate model quantum computer before we have error correction. And then we still need to scale to solve useful real-world problems. And as a result, we're still many years out because no matter what you hear, in the -- from the industry, there are still very hard problems that need to be solved around error correction. It's not just a matter of engineering. And then there are still very hard problems that need to be solved in scaling. It's not just a matter of engineering. And so we do think that it's still a number of years before we will see a scaled error corrected gate model system that is commercially viable.

So for us, though, the focus is on removing the risk. In other words, providing clearer line of sight to being able to deliver that by, a, driving the R&D efforts needed and/or possibly bringing in-house really great things that are going on out there in the industry. And so it's less, I think, about accelerating the time frame and more about kind of being much more concrete on exactly what the road map is that will get us there.

Our next question is coming from Kevin Garrigan of Rosenblatt Securities.

I'm going to switch gears a little bit. And you mentioned you signed renewing customer engagements. Can you give us a sense of what your retention rate is? And the customers that do renew, have they typically already had another application in mind that they wanted to use your quantum annealer for? Or does the team show them how else they can benefit from quantum and that gets the renewed engagement?

Yes. John, do you want to talk about the retention rate, and then I can provide color on kind of how we grow with our customers?

Sure. On average, our retention rate going back over approximately like a 4-quarter period is in excess of 90%.

So we do have a very high retention rate. But we've talked a little bit about this in the past. You need to keep in mind that we have 2 different types of quantum compute-as-a-service customers. There are customers that we call do-it-yourself, where they come in and they buy some access, a developer seat and start kind of exploring, doing research, trying to develop applications on their own. And these customers tend to just continue to renew quarter after quarter or year after year, but haven't been kind of growing or converting from experimentation to actual production applications.

And so one of the things we are focused on is how to help them move faster into production. But not all of them are even kind of at the stage where it makes sense to do that. I mean, some of them are research organizations, which would never convert to production applications. And some of them are smaller organizations that really are just experimenting. So rather, it's kind of really understanding who those do-it-yourself customers are and focusing on trying to engage them with kind of help to move forward. And in some sense, the Quantum LaunchPad program was put in place in part specifically to do that to move those customers off of just do-it-yourself into the LaunchPad program where it comes with some support from our professional services team.

Then the other class of customers is the customers that have engaged us through a professional services engagement. And those are the different customers that, as John said, it's much larger customers that are engaging us now with much larger projects. For example, that aerospace company, I mean, a Fortune 500 company engaged us on 12 different applications from the outset. And we did the evaluation and are now working on the first of those applications with a plan to move through all of them and then take it to other aerospace companies.

So our approach now really is to work with these larger companies to really kind of find a flagship customer in each industry or vertical area, work with them the way we did with this customer. And then as we're helping them progress through the applications, take that out to other customers in the same industry and so on.

Our next question is coming from Craig Ellis of B. Riley Securities.

I wanted to follow-up on my first question but take it in a different direction. So clearly, there's a lot of business model flexibility you now have with a much higher cash balance. And one of the things you can do is, is organically invest a greater amount in R&D and marketing. What I'd like your help on is understanding how you're evaluating current intensity versus higher levels and what we should expect is you evaluate where you can go with internal investment to accelerate your path to further commerciality, especially with this growing Quotient Fortune 500 Global 2000 customers in your pipeline.

Yes. So Craig, first of all, we have in the past said that we are investing in go-to-market. We made a significant, for us, investment in the sales portion of go-to-market sales and technical account management as we over doubled the size of that team in the first half of this year. And now as John said, we're focused on building out the professional services team in support of that. But we're also taking it a step at a time. So we've made an investment and the pipeline looks good. We're making progress. It's taking a little bit longer to get these deals closed than we had expected because of the size of the customer and the complexity of the processes involved, but we're making good forward progress. And we want to see that we're getting a return on that go-to-market investment. And then once that's been validated, we'll continue to grow there.

On the R&D side, we also are starting to make some additional investments in R&D. We talked about the advanced cryogenic packaging work. We've talked a little bit about quantum AI. These are areas where we are starting to make some incremental investments. And beyond cryogenic packaging and its impact on both annealing and gate, there will be incremental investments on the gate model side as well as we continue to kind of work through all the technical elements and R&D elements of that program. And John kind of gave you a metric to think about with respect to an increase in spend throughout the remainder of this year.

We also have a follow-up question from Richard Shannon of Craig-Hallum.

So I had noticed that Triumph had mentioned using Advantage2 in their upcoming research. Are you in talks with them for a sale of QPU? And is this someone who you have recently been talking to or you have mentioned that you had been talking to about sales?

So Triumph is actually working with our system today, and they have seen really good results leveraging our system to do basically generative AI around a particle acceleration problem that they're dealing with. And we've talked about them in the past. We've talked about the work that they're doing. I mentioned them a little earlier. So they are a customer. They are working with our system. They are working with our system in the area of generative AI. They have seen some really good results with that, and we're continuing to grow that relationship with them.

As far as the system sale, I haven't really talked about who is in the pipeline and who we're engaged with other than we have now begun talking about Yonsei University in South Korea.

Appreciate the color. Congrats.

[Operator Instructions] There are no further questions at this time. I would now like to turn the call back over to Alan Baratz for his closing remarks. Please go ahead.

Thank you, operator. So as I think you all know, at D-Wave, our mission is to help customers realize the value of quantum computing right now. Our second quarter results show continued progress in service of that mission across R&D, go-to-market, customer application development and more. Everything we build is designed to provide lasting value for our customers and shareholders, and the future looks very bright. So thank you all for taking the time to join us today.

Ladies and gentlemen, this concludes today's conference call. Thank you for your participation. You may now disconnect.