Opening Keynote | Qubits Europe 2026

Thank you, everyone. I do want to assure you that I am Alan Baratz. I have no clue who the guy in the picture is. It is a real pleasure for me to have the opportunity to be here with you all today, and I thank you all for taking the time to be with us. So, it was Richard Feynman who coined the term quantum computer back in 1981. And there are quite a few different interesting, uh, quotes attributed to him, but one that I think is particularly relevant and important today is the quote on this chart: If you think you understand quantum mechanics, you don't understand quantum mechanics. Now, what's so significant about that is that quantum mechanics and quantum computing are very complex concepts and, frankly, challenging to understand. Unfortunately, too many in our industry are using that to spread misinformation and hype, knowing that it will be impossible for most people to separate hype from reality. What I'm hoping to do here today is to spend a little bit of time helping to try to find the signal in all the noise that's being generated in the quantum computing industry.

So, let's start by talking about what is quantum computing. I'm sure everybody in this room has heard one definition or another. Um, you know, I'm sure you've heard, well, you know, classical computers use bits and a bit can be either a zero or a one. Quantum computers use qubits. Qubits can be zero and one at the same time. And that's a really important quantum mechanical property. But I want to give you a slightly different take on what quantum computing is all about. For me, quantum computing is energy-efficient computing for solving hard computational problems. Think about that for a minute. Energy-efficient computing. In a world where the biggest issue we're all facing and hearing about with respect to computing is energy consumption. That's a pretty important statement. So, where does that energy efficiency come from? It comes from using quantum mechanical effects to perform computations much faster than they can be performed classically and thereby using a lot less energy, a lot less power to perform those computations.

Now, there are a number of different technologies that are being used today to develop quantum computers. But at a higher level, there are two very different and both very important architectural approaches to quantum computing. One is called annealing and the other is called gate. These are very different approaches to quantum and they are each good at solving different types of problems. And that concept is really important to us at D-Wave and what makes us quite unique in the industry. And I'm going to talk a bit more about that as we move forward. But what I thought I'd do is just take a minute to explain these two different approaches because they aren't often talked about. Part of the reason why they aren't often talked about is because there's pretty much only one company that is really focused on and bringing annealing systems to market. That is D-Wave. Uh, pretty much everybody else in the industry is focused on gate. D-Wave now as well. But as a result of that, you don't hear a lot about annealing and you should because it is critically important and there are problems that require it if you want to see a quantum speedup. So, okay, let's talk about annealing versus gate. And let's start by talking about classical computing. Conventional computers.

Uh, suppose I have a problem of trying to find the lowest point in a landscape as represented by the landscape shown in the three pictures where the lowest point would be the valley to the far left. This is a really interesting problem because if you expand it to three dimensions and multiple dimensions, this is one of the hardest computational problems that we deal with. It's known as an exponentially hard problem. Namely, the amount of time required to compute the solution to this problem grows exponentially with the size of the problem and even when you get to small problems, it can be outside of the reach of our fastest supercomputers. Okay, so how does a classical computer solve this problem? It's kind of like you take a tennis ball and you drop it and it hits a point. That's one possible solution to the problem. And the way classical computers work, because they're based on bits and bits can only be a zero or a one at any given point in time, at any point in time, the classical computer is looking at that one possible solution and trying to figure out where to go next. So, you know, we have algorithms that are instructions to the computer to help it figure out where do I search next? Well, maybe I'll try to go downhill. Right? So, let me look at the next point and the next point and the next point. And all classical computers and classical algorithms work that way. At any point in time they're looking at one possible solution and trying to intelligently figure out is there a better solution and if so, what might that be? Now they're looking at that solution and trying to continue working their way toward the optimal solution. The problem with this use case is that there's not enough structure in the problem to have efficient algorithms. That's why it's an exponentially hard problem. Okay, that's classical computing. We see one possible solution at a time iterating our way to the optimal solution.

Let's look at gate model systems. Gate model quantum computers are like dropping an infinite number of tennis balls. So, now the computer can see all the possible solutions, but it's got to figure out which one is the lowest point, which is the optimal solution. And it does that by performing operations on either all or subsets of those possible solutions trying to figure out which is the best one. And, and it literally does that. I mean, operations like, you know, addition, multiplication, exponentiation, logical operations, but not on a single possible solution, on collections of solutions all at the same time. And that's where some of the real speed comes from in these systems. Okay? One thing to keep in mind though, what we're doing is applying digital computation to quantum mechanical systems. Quantum mechanical systems weren't really designed to do digital computation. So, we're kind of fighting against nature when we do this. And as a result, errors get introduced rapidly. And they build up rapidly. And so, these gate model systems are very, very hard to control because of the fact that we're fighting against nature, because of the fact that nature pushes back with errors, and because of the fact that if we can't correct these errors, we can't perform a computation. But we're working on error correction, and I'll talk a bit more about that a little later.

Annealing is very different. Annealing is like, you know, rain on the landscape and see where the water pools. That's pretty much the way annealing quantum computers work. And that works with nature. That works with quantum mechanics. Right? We use the fact that quantum mechanical systems want to be in a low energy state. They want to seek out that low point like water does. Right? And that's really what annealing is all about. So, you can see these are three very different approaches to computation, and they are each good at solving different types of problems. So, you know, when you hear about quantum computing, keep in mind, right? Mostly, you're going to hear about gate model systems, cuz that's what almost everybody in the industry, including D-Wave, is working on. Occasionally, you'll hear about annealing, cuz we're focused on that. And the thing to keep in mind is that there are problems that require annealing. For example, this particular problem, it's an optimization problem, it requires annealing. You can't get a speed up on gate model systems for this problem. That's been mathematically proven. So, gate model systems have their limitations. They're good at some things like quantum chemistry, materials discovery, not very good at optimization. Annealing, really good at optimization, not very good at quantum chemistry or materials design.

Okay, but whatever approach to quantum, it is important. There are many, many, many problems that either are out of the reach of classical, you can't even start to solve them without quantum computers, things like designer drugs, global weather modeling. And then there are many, many problems where, you know, we are solving them today with classical computers, but they're so hard that we're using, you know, heuristics, and we're simplifying the problem to try to come up with what we hope are good enough solutions, but quantum computing can allow us to get to optimal solutions faster. And that's really important, especially when you're talking about problems that you're using to run your business. Okay? So, quantum computing is becoming extremely important. That's becoming more and more well recognized. Pretty much every country around the world is now focused on ensuring that they have a quantum program, that they are able to get access to quantum computers when needed. Um, here in the UK, there is a 2 and 1/2 billion pound 10-year, uh, investment in, in quantum strategy, and work going on to put other programs in place. Germany has a 3 billion euro quantum technologies framework. The European Union has EuroHPC which has a quantum program. They're working on their new quantum act again to drive funding. So, there is a fair amount of money going into quantum from most governments around the world. But, remember this. The country putting more money into quantum than any other country is China at 17 billion US dollars. United States is essentially second at about 5 to 6 billion. Okay? So, we have a ways to go. All right?

In addition, there is absolutely growing customer interest in quantum computing. Now, it's taking time and part of the reason why it's taking time is until recently, we did not even have quantum computers that were large enough and powerful enough to make a business impact, to actually solve business problems and deliver an ROI. Today, we do have that. Our annealing quantum computers are doing that today and I'm going to talk more about that. Gate model systems, not there yet. We still need to solve that error correction problem. We still need to scale them. It'll take a few years before we get there on gate. But, we are there today and we're starting to see companies embrace it. Companies like BASF who you'll hear from a little later or Ford Autosan, Optimark, you'll also hear from a little later. Even North Wales Police here in the UK. They've used it to optimize forward placement of police officers. What they found was leveraging the quantum system, they can compute placements that reduce the average response time by about 50%. Okay?

The thing though we need to watch out for is that we don't stifle inflation. I, sorry. Stifle innovation. I have no clue where that came from. Uh, that we don't stifle innovation by becoming protectionists. And we're starting to see that. I mean, we are starting to see countries saying, you know, "I'm only going to invest in companies that are in my country." Right? That starts to become dangerous because we can't share technologies, uh, information. We can't evolve as rapidly, innovate as rapidly. And maybe you even miss out on important technologies and the ability to leverage those technologies. For example, if D-Wave is a US company and we're the only company in the world that has annealing quantum computers, and you're saying, you know, I, you know, I'm in Germany and I'll only invest in German companies, you, you, you can't use annealing. Right? So, we just need to watch out for that.

Okay. So, I said that there's a lot of noise, a lot of hype, a lot of misinformation. And I want to spend a little bit of time trying to talk about how we find the signal in all of that noise. And for me and for us at D-Wave, we believe that there are three things that are important as you think about quantum computing and working with a quantum computing provider. Uh, one, being able to address the full market, the full set of use cases. Two, being able to do that with viable commercial products that three, deliver a meaningful ROI. These are the things you want to be looking out for as you're thinking about, okay, how should I get involved with quantum? Where should I go to get involved with quantum? So, I'm going to spend a little bit of time now talking about each of those three things, starting with addressing the full market for quantum. There are a number of different market studies on quantum, uh, you know, Bain, McKinsey, Boston Consulting Group. This is data from Boston Consulting Group. Uh, I'm using this because almost everybody in the quantum industry goes back to this study. Uh, BCG first published a study about 5 years ago, if you can believe it. They refreshed it a little less than a year ago. But, BCG puts the market for quantum at a little bit less than a trillion dollars in roughly the 20-year time frame. That's a huge market opportunity for quantum computing. That's why there's so much interest, and frankly investment, in quantum right now. BCG also divides that into four different technology or use case areas, shown here. Crypto, optimization, machine learning, and simulation.

Okay. Back to the discussion a few minutes ago with respect to annealing versus gate, annealing is required for optimization. Gate model systems cannot deliver speed up on optimization problems for some very important technological problems, frankly having to do with error correction. Okay. At the same time, annealing systems are not great at crypto and simulation. Okay. So, annealing for optimization, gate for crypto simulation. Either annealing or gate can address AI and machine learning. They do it in slightly different ways, and I'll talk a little bit more about that in a minute. But, the market is huge. In order to address the full market, you need both annealing and gate. D-Wave is the only provider in the world that brings both annealing and gate model systems to market. In addition, if we take just that optimization portion, which annealing can uniquely address, it's huge. Just the optimization portion is a quarter of the total market. Call it 200 billion in the 20-year time frame. Think about that for a minute. That's more than the global cybersecurity market or the semiconductor equipment market. In and of itself, it's a huge market that frankly only annealing can address. So, sometimes you'll hear people say, "Oh, annealing, that's niche." I don't know. Right? The global cybersecurity market, I'm not sure I'd call that a niche market, but that's what we're talking about here. So, huge market for quantum across many different use cases, both annealing and gate required to address all of those use cases, and even just optimization is a very significant portion of that market.

Okay. Commercially viable products. So, you are going to hear everybody in the quantum industry say, "We are commercial today." Well, okay. Uh, how are your products being used? Which customers are using them? What part of their business are they using your system to support? Right? Well, you know, um, customer X is doing experimentation with our system, and they've, they've able to show that on this particular problem, our quantum computer can solve it better than a laptop. We're commercial today. Okay, that's the problem. The term commercial, in my view, is being grossly misused. This is part of the hype, right? Part of the misinformation, part of the misleading nature of the industry. Let's talk about what is commercial. For me, commercial means the product is being used as a part of a company's business operations. Not research experimentation. It's an integral part of their business operations. That's commercial. Okay?

At D-Wave, our annealing systems have achieved that milestone. We've got a complete suite of products in the annealing arena, from our quantum computers, Advantage 2 is our latest generation system with, uh, almost 4,500 qubits. Uh, it's accessible through our Leap Quantum Cloud Service, designed not just to support research experimentation, but to support business applications in production, building in reliability, availability, security, privacy. We've passed multiple audits on it. We have SOC 2 Type 2 compliant, and we offer service level agreements. Okay? Uh, a complete suite of software developer tools, hybrid solvers to be able to use quantum and classical together, all the way through to services to help understand which applications can most benefit from quantum and how to build and deploy those applications. And we've got a number of companies that are using our systems today as a part of their commercial operations, truly commercially viable.

In addition, our annealing quantum computer, Advantage 2, is the only quantum computer on which quantum supremacy has been demonstrated on a real-world problem. Not a contrived problem, but a real-world problem. This is related to computing properties of materials, where we were able to compute these properties in 20 minutes. It would take nearly a million years on the fastest supercomputers in the world. We did this work with researchers around the world leveraging some of the fastest exascale computers to get the scaling measurements. This is a true supremacy result on a truly useful problem. We published it in Science a year ago. We published it on the archive 2 years ago. There have been attempts to undermine it, but it stands 2 years later.

We have a roadmap for our annealing systems. Uh, we've announced that our Advantage 3 system will have 100,000 qubits. Think about that. We're going from 4,500 in Advantage 2 to 100,000 in Advantage 3. You might say, "Alan, how, how do you make a jump like that in one product generation?" The answer is multi-chip. Up until now, with each generation of our annealing quantum computer, we basically put more qubits on a chip. With Advantage 3, we will be interconnecting the processor chips. So this will be our first multi-chip solution. There are two, two hard problems that need to get addressed in order for us to be successful in that. First, we need to be able to interconnect those processor chips with superconducting bonds. We use superconducting technology that preserve quantum mechanical properties across the chips. Two, we need to be able to scale the system without having an explosion in the control electronics and the IO that has to talk to the chip from room temperature. Okay? So, on the first one, we've been doing work with the NASA Jet Propulsion Lab, and we actually have a process up and running today to provide that interconnect. So, we're making good progress on being able to interconnect the processor chips. On the scaling, we've already made great progress with that even on our Advantage 2 system. So, we control our 4,500 qubit Advantage 2 system today with only about, uh, 2 to 300 IO lines. Now, think about that. Everybody else in the industry requires three to five IO lines per qubit. If we needed three to five IO lines per qubit, our 4,500 system would require 40, 4,500 qubit system would require 12 to 20,000 IO lines. We control it with 2 to 300. That's because we have what we call cryogenic control on-chip control technology. Data multiplexing, pipelining, addressing, so we don't have an explosion in IO. As we go to 100,000 qubits, we, that's 25 times the number of qubits over 4,500, we don't want 25 times the IO line. So, we had to modify our control architecture to be able to control the 100,000 qubits still with only 2 to 300 IO lines, and we've made great progress on that as well. In fact, we've just created the masks for our first next generation control chip that we can fabricate and start experimenting with that. So, roadmap for annealing, uh, on the path to 100,000 qubits.

Okay, let's switch gears and talk about gate model. So, uh, in January, we announced the acquisition of a company that's spun out of Yale University called Quantum Circuits that has absolutely revolutionary gate model qubit technology. I'm going to talk about that in just a minute, but when it comes to gate model systems, there are three things that matter. One, you need to be able to efficiently error correct. I talked about error correction at the very beginning. It's critical. And you need to do it and you need to do it efficiently. Two, you need to be able to scale these systems. 100 qubits isn't going to get you there. You need 10,000 or more qubits to do anything useful. And third, you want fast systems. Nobody wants a slow quantum computer. Okay? Um, now, fast and efficient is kind of interesting. There's a little bit of a debate going on in the quantum industry right now around which underlying technology is best for building a gate model quantum computer. Is it superconducting, which is what we use for both our gate model and our annealing systems? IBM uses it, Google uses it, Rigetti uses it, or is it trapped ion? What IonQ or Quantinuum use? Or is it neutral atom? What Qera or Pascal use? Or is it photonics? PsiQuantum. What's the best? So, how does the debate unfold? Well, it unfolds as follows. The superconducting folks, like us, will say superconducting compute speeds are a thousand times faster than ions or atoms. And they are, and that's physics. Hardware and software isn't going to change that. The gate speeds are a thousand times faster with superconducting systems. Well, on the other side, they'll say, "Okay, you may be faster, but atoms, ions are natural qubits. Lower error rates, much easier to error correct. So, we're going to get error corrected systems way before you superconducting folks, if you can even ever get there, because your qubits are very noisy." So, that's the debate, speed versus fidelity. Right? Okay. Keep that in mind, because we just changed the game on that debate.

The technology that Quantum Circuits had developed that we acquired in January is a revolutionary qubit technology. It's called dual rail qubit. And what's so significant about this technology is the qubits themselves can detect errors. It's inherent in the qubit. It can detect 90% of the errors that occur. Well, the implication of that is that you can do error correction much more easily on those qubits, much more efficiently on those qubits, but they're superconducting. So, they're a thousand times faster than ions or atoms, but they're just as easy, nothing's easy, but you know, they're just as efficient to error correct as atoms or ions. So, now we have technology that is fast and efficient. I think it's game over for atoms and ions, but I am a superconducting proponent. Of course, you also need to scale. Right? So, how we going to scale these systems? Remember, I said everybody else in the industry requires three to five IO lines per qubit. Right? Well, on our annealing systems, we're controlling 4,500 qubits with two to 300 IO lines. We are taking that cryogenic control technology and moving it into the gate model arena, so that we will be able to scale these systems efficiently as well. So, we bring together the dual rail technology, which is fast and efficient for error correction, with the cryogenic control technology from our annealing systems to address both of the two hard problems in bringing scaled fault-tolerant gate model systems to market. So, you can kind of think of trapped ions and neutral atoms as a bicycle. They're simple, efficient, but really slow. Whereas superconducting is like an old, you know, piston airplane, right? Really complex, certainly compared to a bicycle, but a lot faster. Dual rail qubits, they're a turbine engine jet, much simpler, maybe not quite as simple as a bicycle, but much simpler than a piston airplane, and very fast.

Okay, let me kind of put some numbers on this for you. Um, what I'm looking at here is some metrics associated with error correction. Now, there are two things that matter when we talk about error correction in a gate model system. One, how fast can we perform the error correction, and two, how efficiently can we perform it? And by efficiency, what I mean is how many physical qubits does it take to create one error-corrected logical qubit? Okay? So, the table on the left has a column called speed. This is the time to run an error-correcting cycle. You can see that for superconducting, on the order of 5 microseconds to run an error-correcting cycle. Trapped ion, 30,000 microseconds. Neutral atom, 200,000. This is that thousand times factor I was talking about previously. This is not something that you just wipe away in

software. This is fundamental physics. This doesn't change. These are much slower systems, ions and atoms, than superconducting. Now, let's talk about error correction. So, let's not worry about D-Wave for a minute, but just look at the others: superconducting, neutral atom, trapped ion. So, I want to introduce a term called lambda. What is lambda? Lambda is... Think of it as bang for the buck with increased error correction.

So, in other words, how much lower is the error rate as you increase the error correcting code? Or, said another way, how much lower does the error rate go as you increase the number of physical qubits per logical qubit? Because when you're increasing the error correcting code, you're increasing the number of physical qubits per logical qubit. So, the graph on the right shows two lines. One line is a lambda of two. For each increase in the error correcting code, you're cutting the error rate in half, a factor of two. The blue line is a lambda of 10. For each increase in the error correcting code, you are basically cutting the error rate by a factor of 10, one-tenth.

All right. So, the best that's been shown in the industry so far, including the Google Willow demonstration, is two. That's kind of where we are as an industry today. There's been some preliminary work done that shows that with trapped ions, maybe we can do better than two, but nothing is yet at 10. The D-Wave dual rail technology, we've got a very early version of the system, small, but the evidence points to a lambda of 10. This is the quantification of what I said before: efficient and fast. That's the dual rail technology.

And then, our gate model roadmap, we announced this two or three weeks ago. We are currently still really in the NISQ era: noisy intermediate-scale, non-error-corrected gate model systems. You can't really do anything with them other than play around. You can't do anything useful before we get to error correction because of the problem I described a few minutes ago. But what's post-NISQ? Post-NISQ is where we have some ability to manage errors. The dual rail qubits give us that because they will tell us when an error has occurred, and we can make that available to algorithms and to developers to use as part of the development of application software. So, that's what we're calling the post-NISQ era. Still not error corrected, but more information that may be useful. We don't know yet, but it may be useful.

And we've announced a roadmap for D-Wave: 17 dual rail qubits in 2026, which will allow us to have a very small distance-three error correcting code. Back to the previous chart, you see 17 qubits, distance three, and experiment with the properties of that. Next year, 49 qubits, a distance-five code. And then in 2028, 181 qubits would be a distance-nine code. The point is, with a lambda of 10, by the time we get to 181 qubits, we should have a three-order-of-magnitude reduction in the error rate, taking us to the point where if we built a system out of this, we could actually support a million gate operations, a computation with a million operations, which we think is the starting point for commercial viability.

Then as we move into 2030, we start scaling. 10 logical qubits, each of them using roughly 181 physical qubits. 10 logical qubits in 2030, 100 logical qubits in 2032. That's really the point at which we have the first truly commercially viable gate model quantum computers: 100 logical qubits able to support up to a million operations. By the way, when other people talk about logical qubits, ask them: how many operations will you be able to support? That's critical. There's too much noise in the system today around logical qubits. People are using the term logical qubit to refer to even a small improvement in the error rate over the physical qubit. We need dramatic improvements. So, ask them: how many operations will you actually be able to support with those logical qubits?

All right. That's a bit on our roadmap: commercially viable products, annealing commercial today, moving to gate model. I now want to spend a little bit of time talking about the actual use of our quantum computing systems. And I'm actually going to have some help doing that. So, first of all, I'd like you to welcome Derek Kosty, who's senior director for platform and data at Optum Rx, a part of UnitedHealth Group. Derek.

Hello, Derek. Thanks for being here. No, we don't get to sit. We're standing. Chairs are for later. Okay, so as I said, Derek, it's a real pleasure to have you here with us. We've been working with you for a little over a year now and some really interesting and exciting things have happened, but let me start with: what was the impetus for Optum to get involved with quantum computing in the first place?

What were the challenges that you were running into using classical to solve these problems?

Yeah, and that's a really important point. A lot of these computationally challenging business problems are being solved today. It's just that they're being simplified and we're using heuristics and trying to come up with what we hope are good enough solutions. And now, let's look at quantum and see if we can solve them better, faster. And you mentioned that you brought somebody new on the team, but how did you actually discover D-Wave?

So, I'd like to say that our crack sales team actually approached you. Guys, get your act together. We got to you first. We got an inbound and that was great. Okay, so maybe you could talk a little bit about the journey that you and your team followed in working with us.

So, you've kind of almost answered, but where does this work stand today?

And you use the word production, and that's back to what I said before: it's not about research experimentation, it's about using it as part of your business.

Can you talk about your roadmap for working with D-Wave going forward?

And just to wrap up, what advice would you give other Fortune 100 companies who have complex optimization problems?

It is real. Derek, thank you. Appreciate your time.

I must tell you I am really appreciative of Derek being here. He shared with me this morning that he's now running off to go to a wedding. He's got to get on an airplane and go to a wedding. So I'm not sure my wife would have allowed me to come do this so close to that. Okay. I said I was going to talk a little bit about AI, so let me do that. I believe that there are a lot of synergies between quantum and AI. I think there are areas in which AI can benefit quantum. There are areas in which quantum can benefit AI.

Just to talk about for a minute quantum benefiting AI, at D-Wave, we have really the first and only true commercial quantum AI application. This is work that we've done with a Japanese pharmaceutical company called Shionogi. Basically, they developed a generative AI model for drug discovery. This is a model that they trained on their molecular data set to generate novel molecular structures for human drugs. They trained the model classically and they were getting okay results. Not a lot of molecular structures generated were well-suited to human drugs, but some were. They then enhanced the model to leverage our Advantage annealing quantum computer. And what they found was they got much more accurate models. And now the molecules being generated, roughly 10 times more of them are well-suited to human drugs across a variety of different metrics.

We now have a number of other companies that are interested in doing something very similar. And in our labs, we are actually doing some work to demonstrate more broadly how quantum can aid in both model training and very importantly inference. And we're seeing some very promising early results about the ability to speed up inference and as a result do it with much less power consumption. Back to that energy efficiency.

And then finally, just a minute on blockchain. When we announced our quantum supremacy result, initially two years ago and then the Science publication a year ago, we also put out a paper on how you could use that computation to basically create a new blockchain proof of work, a quantum proof of work. We basically built a hashing function out of that supremacy computation. And we implemented it on a distributed quantum system using four of our quantum computers, up and running. We're continuing to experiment with it. The interesting thing about that is if Bitcoin was based on that quantum proof of work rather than the current SHA-256, it would consume about a thousand times less power today to do the mining. And this is a quantum-safe computation, unlike SHA-256, which eventually can be broken by quantum.

In addition, we've been working with a small company called Post-Quantum, who has developed a hybrid blockchain proof of work, where you can mine with either classical resources, CPUs and GPUs, or with our quantum systems. Actually, you could use any quantum system, but they're using ours right now. They currently have over 18,000 miners on the testnet. There are over 1,600 nodes, compute engines on the testnet, CPUs, GPUs, and then our quantum computer. We don't put the quantum computer online constantly. We put it online for about five minutes a day. The reason is when the quantum computer goes online, it wins all the blocks, which is kind of what you'd expect. So, this is in part demonstrating the advantage of the quantum computer, and in part demonstrating the ability to do more energy-efficient blockchain proof of work.

Okay. Just to kind of wrap up this portion of it. If we go back to Feynman, there's a lot of noise in the system. And it's because quantum mechanics and quantum computing are very challenging and very difficult to understand. And it's really easy to say things where it's hard for any of us to really know whether they're accurate or not. But we think it really is important to find the signal in the noise.

D-Wave is pretty unique in the industry today in that we're the only dual-platform quantum company bringing both annealing and gate model systems to market to address the full market for quantum. We're the only company that has truly commercially viable quantum computers today. Derek talked about applications in production as a part of a company's business operations, not just research experimentation. And these systems are delivering real customer value.

With that, we're going to do a little bit of a fireside chat. So, I'm going to ask Jeremy Kahn to join me here on stage. Jeremy is Fortune's AI editor spearheading the publication's coverage of artificial intelligence. He is the author of Mastering AI: A Survival Guide to Our Superpowered Future and is the lead author of Fortune's Eye on AI newsletter. Before joining Fortune in 2019, he spent eight years at Bloomberg as a technology reporter and a senior writer for Bloomberg Markets Magazine. His writing has also appeared in numerous publications, including The New York Times, the International Herald Tribune, Newsweek, The Atlantic, Smithsonian, The Boston Globe, and Portfolio. He holds degrees from the University of Pennsylvania and the London School of Economics, and he's currently based in London. Jeremy?

Yeah. So it really gets back to applications in production leveraging a quantum computer. In other words, we really have transitioned from someday quantum computing will deliver value to quantum computing is delivering value today. Customers are no longer asking me when can I use quantum computing? They're asking how do we get started? So while we didn't have the headlines and the fireworks associated with what you might think of as a ChatGPT moment, we have made the transition from quantum computing will someday be important to quantum computing is being used today. It is creating value today. It is improving business operations today.

Within the last year. Yeah. Literally within the last year. Frankly, we're the only company in the world that has quantum computers that are able to support commercial applications in production. That is associated with our annealing systems. Our gate model systems are not there yet. Nobody else's gate model systems are there yet. But even we didn't have that until we brought our Advantage 2 system to market, and that's only a little over a year old. So, it's happened all within the last roughly 12 to 15 months.

Yeah. So, because it's the annealing systems that are commercial today, it really does go back primarily to optimization problems. And I'll talk about some of those in a minute, but also the beginnings of commercial traction in AI. So, in AI, I talked about Shionogi and the first really commercial quantum AI application. And I think we're going to start seeing a lot more of that. In the optimization arena, it's everything from automotive manufacturing optimization to chemical production scheduling. I talked about the North Wales Police and forward placement of officers. So, optimization is pretty broad, right? Workforce scheduling. We've got customers using us for that. But whenever you've got a business problem where you need to optimize some function subject to a bunch of constraints, that's an optimization problem, and annealing can help.

So again, just over the last year. But over the last year, we have gone from helping customers with proofs of concept where the deal size to do that would be small six figures, to larger companies. We now have one of the largest airlines in the world, one of the largest aerospace companies in the world.