Quantum Computing is Real & Commercial Today, says D-Wave CEO

I'm great. Thanks for the opportunity to be here, Will.

Sure. Starting with the second question first: what is quantum computing? It's very simply using quantum mechanical effects to solve hard computational problems faster than they can be solved using classical computers. And these quantum mechanical effects are things like superposition, entanglement, tunneling. All of these quantum mechanical effects are used within our D-Wave quantum computers to help solve important problems. As far as D-Wave and what we do, we are a full-stack quantum computing provider. We provide everything from the quantum computers — we design and manufacture them ourselves — to our quantum cloud service, which is currently the primary mechanism that our customers use to access our quantum systems to run their applications, to a complete suite of software development tools, all open source, so that customers can basically help us to evolve and improve the tool set, through to professional services to help our customers build and deploy applications. But what really makes us unique in the quantum industry is that we are the first and only quantum computing company that is actually supporting customers today with applications leveraging our quantum systems in production as a part of their business operations. And so these include companies like Pattison Food Group, NTT Docomo, or Ford Otosan that are all using our quantum systems and services today as a part of their business operations.

That's correct. So we design and manufacture and make available the quantum computers. Our current generation system is called Advantage. It is a 5,000-qubit quantum computer, the largest and most powerful quantum computer in the world. And as I said, we have a complete suite of software development tools that allows our customers to build and deploy applications that leverage those quantum systems. And I should also point out that those software development tools include access to what we call hybrid solvers that are part of our quantum cloud service. And so this is where we use both classical computing CPUs and GPUs together with the quantum computer to solve hard computational problems, where we're basically allowing each processor to solve the part of the problem that's best suited to the capabilities of that processor.

Well, first of all, let me point out that our quantum computers are much larger than any other quantum computers that are being developed today. All of the other quantum systems are in the tens to a few hundred qubits. So our systems are an order of magnitude larger than any other quantum computer. Now, to put it in context versus classical computing, let me do it in terms of a specific application. We recently published in Nature a demonstration of our quantum computers solving a magnetic materials computation in minutes that would take nearly a million years to solve on Frontier at Oak Ridge National Lab, which is one of the largest and most powerful supercomputers in the world. So at a few thousand qubits, our quantum computers are performing computations well beyond what's possible with classical computers, even the most powerful supercomputers in the world.

Well, we like to think about it that way. Yeah. Supercomputers are last year's technology. They're very capable. There are very important problems that can be solved on classical computers, whether they're CPUs or GPUs or massively parallel GPU systems. But there are also problems that cannot be practically solved on those systems. And that's where the next wave of quantum computing comes in. And I should make another point as well, which is that it's not just about the time to solve the problems. It's also about the power consumption. So as I think we're all aware now, GPUs are massive power hogs. And you know, we hear about it every day that big tech is looking to buy nuclear power plants to kind of power these hyperscaler compute infrastructure environments. But quantum computers are much more energy efficient. So that computation I referred to a few minutes ago in the area of magnetic materials, it took us about 12 kilowatts to perform that computation. If we were to perform it on Frontier, it would take well over global annual energy consumption. So this is a much more energy-efficient approach to computing as well.

So that's really in the way that quantum computers operate. As I said early on, quantum computing is all about using quantum mechanical effects to solve hard computational problems. And let me give you a very high-level explanation of what I mean by that. If you think about current computers, information is stored in bits. And a bit can be either a zero or a one at any point in time. So what that means is that at any point in time, a classical computer is evaluating one possible solution to the problem. And the way classical computers work is they basically very intelligently iterate from one potential solution to a problem to the next, searching for the best possible solution. And algorithms are the term associated with that intelligent search process. Quantum computers use what we call qubits and superposition. And what this means is that at any point in time, qubits can be combinations of zero and one. So what that means is that at any point in time, a quantum computer is able to evaluate and operate on multiple possible solutions to the problem at the same time, allowing it to iterate its way to the best possible or correct solution much faster. And that's the way our quantum computers work.

Yeah. So there are two main architectural approaches to quantum computing. One is called quantum annealing, as you just mentioned, and D-Wave is actually the only company in the world that provides annealing quantum computers. Everybody else is focused on gate-model quantum computing. Now, there are some important differences between the two. Perhaps one of the most important differences is that annealing is a much easier technology to work with. It's easier to scale and it's much less sensitive to errors. And as a result, our quantum computers are an order of magnitude larger than any other quantum computer. And they are able to deliver good, if not optimal, solutions to hard computational problems today without the need for error correction. On the gate-model side, the technology is much more challenging and it's very, very sensitive to errors. In fact, there's no evidence at all that a gate-model quantum computer will be able to solve a useful commercial problem without error correction. And we're still many years away from error correction on gate-model systems. And that's why D-Wave is commercial today. That's why we have customers that are using our systems as a part of their business operations today. And that's just simply not possible with gate-model systems because they're still at a very immature R&D stage where really all that's going on is research experimentation.

So there's a really interesting historical answer to that question. D-Wave began building a quantum computer over 15 years ago. We're the first company in the world to start building a quantum computer. At that point in time, the science, the engineering had not yet progressed to the point where there was line of sight to how you could build a gate-model system. But there was line of sight to how you could build an annealing quantum computer. And so we decided to start with annealing quantum computing. In some sense, it's the only thing we could do at that point in time. Now, we knew that annealing would be very good at solving business optimization problems. Frankly, most of the important hard problems that businesses need to solve are business optimization problems. And we knew that annealing would be good at that. But we also knew that there were things it could not solve. For example, you can't use an annealing quantum computer for molecular discovery, for quantum chemistry, for example. So we decided to start with annealing. Eight or nine years ago, when everybody else decided to seriously pursue developing a quantum computer, the science and the engineering had progressed to the point where it was believed you probably could build a gate-model system. And at that time — and this is a really important point — at that time it was believed that a gate-model quantum computer could perform all computations, could basically solve all types of problems better than classical. And so everybody else sort of adopted the point of view: well, we know that annealing is good at some things, but not all things. We believe gate model is good at all things. We now think we know how to build a gate-model system, so let's go there. Now, fast forward to a couple of years ago. And this is a really kind of interesting twist of fate, at least for D-Wave. A couple of years ago, it was proven by researchers in both the US and Europe that gate-model quantum computers will likely never deliver a speedup on optimization problems, that you basically need an annealing system in order to address optimization problems. So now it's not that one is more general than the other. It's not that at all. It's that there are some applications that will always run best on annealing, and there are other applications that will always run best on gate. So you're always going to need both. But because of this kind of interesting twist of fate, D-Wave is really the only company that's doing annealing and can address that, frankly, very important portion of the market. So it's actually worked out quite well for us.

So the primary way that we and our customers access our quantum systems is through our cloud service. Now, we have recently sold one of our current-generation Advantage — we call our current systems Advantage — one of our current-generation Advantage quantum computers to the Jülich Supercomputing Center in Germany. And we did that because they have an interest in tightly integrating our system with their new Jupiter exascale supercomputer. This is the first and only exascale supercomputer in Europe, and they want to tightly integrate it and explore a variety of workflows and use cases in the areas of AI and optimization. And in order to do that, they actually needed to own a system, have it on site, and be able to have control over more of the operating parameters of the system than we make available through our cloud service. But for customers that are simply interested in running their business applications, the cloud service is the way to go. It's a much more cost-effective approach to accessing our quantum systems. So we actually now have two components to our go-to-market model. One of them is the cloud-based service. We call it quantum compute as a service. And that's ideal for commercial companies that are just simply interested in running their business application. And the other is system sale. And that's more for supercomputing centers, government labs that are really interested in tighter integration with other systems and controlling more of the operating parameters of the quantum.

So we actually do not sell applications. We sell access to our quantum systems. We sell our quantum systems, and then we sell professional services to help our customers build and deploy applications that are of value to them. And so as a result, we've helped many companies evaluate applications with respect to what kinds of benefits they could achieve by leveraging our quantum systems. And then for those where they see a sound ROI on the application, we've helped them actually build and benchmark and deploy those applications. So for example, Ford Otosan, which is a joint venture of Ford Motor Company and Koç Holdings in Turkey, we worked with them to develop an application for scheduling the assembly of automotive bodies. And what they found was that by using our quantum computers, we could significantly reduce the time to do the scheduling, which meant that they could adapt in real time to changes throughout the day. And that was important to them. It's now moving into production at Ford Otosan.

So they don't. The applications that are in production today at our customer sites are basically using quantum systems to solve hard optimization problems. However, we are working on ways to integrate our quantum systems within AI environments, and there are two areas that we're looking at. One is basically using the two capabilities together, letting each solve the portion of the problem that it's best at solving. For example, you might use AI to predict future product demand and then use quantum to optimize the supply chain to meet that demand. Here we're using AI and quantum together, each addressing a different portion of the use case, the portion that it's best at addressing. The second area that we are working on, and this is potentially very, very transformative for AI, would be to insert the quantum computer into the AI model training and inference workflows. Why would you do that? This is back to my comment earlier about energy efficiency. So what's the biggest issue right now with AI model training? It's power consumption. However, what we've started working on — and we have early evidence that this could be a very transformative capability — what we've started working on is using the quantum computer to perform some of the computations in model training together with the GPUs, where the two working together can use dramatically less electricity to perform the model training. And we think that this could be very transformative for the AI industry, and we're quite excited about it.

Oh, no. We actually have work going on in that arena. We've actually developed some of the early capabilities. We've used it on smaller data sets and we're seeing good results. We're now in the process of basically starting to work with much larger data sets to essentially prove out the capability and the value.

Yeah. So first of all, let me take just a brief step back and provide some context for all the commercial work that we are doing. So there's a lot of confusion in the marketplace with respect to whether quantum computers can actually deliver value today on important real-world applications or not. Unfortunately, too many of the quantum computing companies are dramatically overstating what they are capable of delivering in that arena. So what we wanted to do was to help customers separate the hype from the reality. And as a result, what we did was we developed a framework that we call Quantum Realized. Basically, what you should look for in a quantum provider if you really want to have confidence that you're going to be able to get value. And there are three components to that framework. First, have the systems provided by the quantum company been able to solve problems better than they can be solved classically? Simple thing, but the truth of the matter is you probably won't find a gate-model quantum computer where you can affirmatively answer that question. But for D-Wave, the problem that I described to you a little while ago in magnetic materials — this is an important and useful problem that we are solving in minutes that essentially can't be solved classically. So, does the vendor have quantum computers that can solve a problem better than it can be solved classically? Second, are the systems from the vendor reliable and available enough for you to actually be able to depend on those systems as a part of your business operations? And in our case, when you access our systems through our quantum cloud service, we have 99.9% availability. And in fact, we offer service level agreements to our customers. We believe we're the only quantum company in the world that actually offers service level agreements. And in fact, if you go out and look at the public data around reliability and availability of other quantum systems, you'll see things like 30% availability, 70% availability, one-week queuing times before a problem will be run. I mean, our problems run in real time with very high availability. So that's the second point. And then the third point in the framework is: is the quantum company actually working with customers on real commercial applications that are in production? And you know, we've talked about Ford Otosan, for example, which has an application on auto body assembly in production today. So those three elements we think are what any company should be looking for and asking about when they start talking to a quantum company. Can your systems solve problems better than they can be solved classically? Are your systems highly reliable and available? And do you have examples of where you've worked with commercial companies to be able to deliver value? And we can answer yes to all three of those questions.

Now, some examples. I talked about Ford Otosan. Another example is NTT Docomo. We've worked with NTT Docomo on optimizing cell tower resource utilization, and there are two elements to the results. One, we've been able to perform this optimization computation in 40 seconds, whereas it had been taking them 27 hours. And two, leveraging the solutions that come out of our quantum systems, each cell tower can support up to 15% more cell phones, which means lower infrastructure costs for NTT Docomo, which is a significant benefit to them. Another example is Pattison Food Group, a Canadian grocery chain, which is basically leveraging us for two different applications. One is for workforce scheduling, taking a computation that was taking them 25 hours a week down to minutes a week. And then last-mile routing for e-commerce delivery. So these are some examples of customers that are actually using our systems today in production. There are other customers that we are working with. For example, BBVA. We're working with them on financial portfolio optimization leveraging both AI and quantum computing. We've talked about some work that we did with the pharmaceutical division of Japan Tobacco, where we've been able to help them enhance their large language model training in a way where they are now able to basically compute better new molecular structures than with the models that were developed without quantum.