Thank you all for taking the time to be here with us this afternoon. We're quite excited to be here and to have the opportunity to tell you a little bit about our technology, our products, our customer use cases, and to actually give you the opportunity to hear from some of our customers themselves. And we're looking forward to this being only the first of many events here in Japan as we look to the future and try to make this a regular event. So with that, why don't we go ahead and jump in.
The last year has been actually quite amazing and quite exciting for the entire quantum industry, both from a commercial perspective as well as from a technology and product perspective. From a commercial perspective, we've seen the first quantum applications go into production. What I mean by that is companies actually using quantum systems as a part of their business operations. And I'm really proud to say that that is based exclusively on D-Wave quantum computers.
It's also projected that across the entire industry, as many as 50 quantum computers will be sold this year. With respect to the independent publicly traded quantum companies, over the course of the last year, their combined market capitalization has increased by almost 1,000% to now over 22 billion dollars, and over double the number of financial analysts are now covering quantum companies as was the case a year ago.
From a technology and science perspective, we have finally, for the first time, seen the ability for a quantum computer to solve an important, useful, real-world problem that cannot be solved classically. And once again, I'm really thrilled and proud to say that that was done on the D-Wave quantum computer.
This is what the entire quantum industry aspires to at this time, and we're the first to actually be able to demonstrate it, and I'll talk more about that in just a few minutes. But there have been other important technological advances in the quantum industry as well. The Google Willow announcement, for example, was critically important as it was the first demonstration of error correction on a gate-model quantum computer. And we've also seen dramatic improvements in qubit and gate fidelities on gate-model systems over the course of the last year. So it's been a truly amazing and exciting year for the quantum industry.
With respect to D-Wave, many of you know who we are, some of you do not. For those who do not, we are a full-stack quantum computing company. We design and manufacture quantum computers. Our current-generation system is called Advantage 2. It is a 4,400-qubit quantum computer and is arguably the largest and the most powerful quantum computer in the world.
Most of our customers access our quantum systems through our quantum cloud service. This cloud service was designed not just to support research experimentation, but to support business applications in production, building in the security, the privacy, the reliability, the availability, and the performance that's required to actually support commercial applications in production. We have over 99.9% uptime and we offer service level agreements. We're the only quantum computing company in the world that has reliability and availability and performance good enough to actually offer service level agreements.
We have a complete suite of software development tools. Our Ocean SDK is all open-source so that developers can help us to improve and extend the tool suite. And we have a services team that is available to help companies understand which of their applications can most benefit from quantum and how to build and deploy those applications.
We have over a hundred customers, most of them commercial, some of them research and education. And we have a very strong intellectual property patent portfolio with over 500 worldwide granted and in-process patents.
And it's also been a pretty exciting year for D-Wave. Some of the exciting things I've already talked about, but we did earlier this year launch our Advantage 2 quantum computer. This is the system on which we were able to demonstrate quantum supremacy. We tried it on our previous-generation system. We were not able to achieve the result, but we were able to do it on our Advantage 2 quantum computer. So a significant boost in performance, which I will also spend a bit more time on in a minute. And as I said, this is the quantum computer on which we were able to demonstrate quantum supremacy, which was published in the peer-reviewed journal Science earlier this year. We then took that result and demonstrated how you could use it to build a hashing function and create a blockchain based on quantum proof of work.
And what's so important about this is not only would this be a quantum-safe blockchain by construction, but if Bitcoin, for example, was mining using this computation rather than SHA-256, Bitcoin mining today would be consuming about a thousand times less electricity. So this is a very energy-efficient platform.
Of course, we've also made progress in a number of other important technological areas. We announced earlier this year that we now have the ability to interconnect processor chips through superconducting bump bonds, which will be critical to our scaling roadmap, which I'm going to talk about in a few minutes. And we announced enhancements to our Ocean software development kit in support of quantum AI, the ability to actually use our quantum computer as a part of the AI model training process.
We sold our first quantum computer. We sold it to the Jülich Supercomputing Center in Germany, where they will be interconnecting it with their Jupiter exascale supercomputer. This is the first and only exascale supercomputer in Europe, benchmarked as the number four fastest HPC system in the world, and they will be interconnecting our quantum system with that supercomputer in support of new AI workloads, materials discovery workloads, and optimization workloads.
And then finally, we introduced a framework that we call Quantum Realized. There's a lot of hype in the quantum industry, and it's becoming increasingly difficult to understand what's real and what's hype. And so we introduced the Quantum Realized framework to help everyone try to get a sense of what they can and can't depend on from various quantum vendors. The Quantum Realized framework has three components.
So when you're thinking about whether to start using a quantum computer, whether to buy a quantum computer, there are three things we think you should keep in mind. One, has the quantum computer actually been able to solve any problem, any problem, better or faster than it can be solved classically? And I'm not talking about faster than it can be solved on your laptop. I'm talking about faster than it can be solved on any classical system. Has the quantum vendor demonstrated this?
You're going to have to search long and hard to find a company that's demonstrated it. Although when I talk about our quantum supremacy work, I'll be pointing out that we have achieved that. Second, are the quantum systems reliable enough that you can count on using them to do useful work?
If it's a one-week queuing time to run a problem, or it takes 100 hours to get a result back, it's going to be hard to use that to do anything useful. Are the systems reliable enough, available enough, and performant enough that you can actually do useful work? And finally, has anybody actually used the systems to do something useful? Are there referenceable customers? We think everybody should keep these three things in mind as they're thinking about whether to start working with a quantum company and a quantum system. Now, yes, that is somewhat self-serving, given that D-Wave is the only company in the world that can actually answer yes to all three of those, or actually even any of the three. But these are important criteria, and we call that Quantum Realized.
Okay. Earlier this year we launched our Advantage 2 quantum computer. It's a 4,400-qubit quantum computer. There are three areas in which Advantage 2 improves upon our earlier-generation Advantage quantum computer. And these are the three areas that we're always focused on with each generation of system. One, connectivity. Advantage 2 increased connectivity from degree 15 to degree 20. So in Advantage, each qubit was connected to 15 others. In Advantage 2, each qubit is connected to 20 others. That's important because the connectivity is related to the size and complexity of problems that we can solve.
Second, coherence time. We're always focused on increasing coherence time because that's related to the time it takes to solve a problem. Longer coherence time, faster time to solution. We doubled the coherence time between Advantage and Advantage 2. And finally, energy scale. Think of energy scale as the precision with which you can specify the problem parameters, kind of like single precision versus double precision. We increased the energy scale in Advantage 2 by 40%. So more connected to solve larger problems, longer coherence time to solve them faster, and increased energy scale to get better solutions to the problems. This is why the Advantage 2 system is so much more powerful than Advantage and why we were able to demonstrate quantum supremacy on it.
And speaking of quantum supremacy, what was it that we actually did? So we were working on the problem of computing properties of materials, specifically magnetic materials. And we wanted to compute properties as the materials get close to what's called a phase transition point. So think of this as water freezing. That's a phase transition point. Magnetic materials go through phase transitions as well. And so we wanted to compute properties of these materials as they get close to a phase transition point. We used several different material structures, four to be specific: two-dimensional lattice, three-dimensional lattice, diamond, and plaquette. We used different size lattices from small to medium to large. And we used different evolution times on the simulation of the materials.
We were able to compute all of these properties for all of the lattices at all sizes and all evolution times in minutes on our Advantage 2 system. For the largest, it would take nearly a million years to solve on the fastest supercomputer in the world. We actually worked with 11 institutions around the world, tens of researchers, including experts in all the classical techniques to solving these problems, whether it be tensor networks approaches like PEPS and DMRG, or neural networks, or quantum Monte Carlo, looking for the best possible approaches and trying to gain the scaling factors so that we could extrapolate to what it would take to solve these problems classically on the fastest supercomputers. And that's the result. We can perform the computation in minutes. It would take nearly a million years classically.
Moreover, it took us about 12 kilowatts of power to perform these computations. It would take on the order of global annual energy consumption to perform them classically. So, not only are we very fast, but we're very energy efficient.
And as we look to the future, our next-generation system is called Advantage 3. Advantage 3 is targeted to be 100,000 qubits. Now, you might say, "Come on, Alan. How are you going to go from 4,000 qubits to 100,000 qubits in one generation?" The answer is easy. Well, easy for me to say, hard for Trevor Lanting, our chief development officer. But it's because for the first time with Advantage 3, we will be moving to a multi-chip system. Up until now, we've grown our systems by putting more qubits on a chip. With Advantage 3, it will be the first time where we will be interconnecting processor chips. And I've already talked about the fact that we've now demonstrated the ability to actually interconnect through superconducting bump bonds.
And if you think about it, going from 4,000 to 100,000 doesn't require that many interconnected chips. 25 interconnected chips, which we'll basically lay out in two dimensions with a tiling structure. So we don't even think we'll need much of an increase in refrigerator size. Maybe a 20% increase in the refrigerator size. We don't think we'll need much of an increase in I/O. We're actually quite unique and very good at I/O for quantum computers. We control today 4,400 qubits with 200 I/O lines.
Think about that. 4,400 qubits with 200 I/O lines. If you go ask any other quantum computing company how many I/O lines it requires per qubit, they will say a few. Anywhere from three to six I/O lines per qubit. If we were using that brute-force approach, our 4,000-qubit processor would require tens of thousands of I/O lines. That's not scalable.
We control 4,000 qubits with 200 I/O lines. We think our 100,000-qubit system won't require more than about 300 I/O lines. So this is well within reach for us. And then we're continuing to make progress on our gate-model quantum computer. So to be clear, everything that I've talked about up until now with respect to Advantage, Advantage 2, Advantage 3, that uses an architecture we call annealing.
Those are annealing quantum computers, and they are very good at solving certain classes of problems. They are excellent at solving optimization problems. They can solve AI machine-learning problems. They can even do factorization, but not using Shor's algorithm, using a different approach. But they cannot solve differential equations problems. You would not use an annealing quantum computer for quantum chemistry, for example. For that, you need a gate-model quantum computer, which is what pretty much everybody else in the world is trying to build. We're the only company that does annealing quantum computing. Everybody else is pursuing gate-model. We are pursuing a gate-model system as well. And the reason is because we have customers with use cases that require both.
For example, think about a drug company that's designing a new drug, putting it through trials, manufacturing, shipping it out. The quantum chemistry up front does require a gate-model quantum computer. Optimizing the trials, optimizing the manufacturing process, that requires an annealing quantum computer. And when I say requires annealing, I actually mean that because there's now a body of mathematical evidence that gate-model systems will likely never deliver a speedup on optimization problems. So we have a bifurcation in the application environment for quantum. We have applications that will always require annealing, think optimization. We have applications that will always require gate, think differential equations. And then we have applications that can be addressed by either.
But we are also working on a gate-model program. And that program is really oriented toward a scaled, error-corrected gate-model system. We are not believers in non-error-corrected gate-model systems. We don't actually think you'll be able to do anything practical or useful with them. So, our focus is on error correction and moving as quickly as possible to a scaled, error-corrected gate-model system.
And as I've already said, our systems are very highly reliable, very performant. We even have SOC 2 Type 2 compliance on our cloud service all the way back through to the quantum computer. We're the only company that has SOC 2 Type 2 compliance all the way through to the quantum computer. And as I said, we do offer service level agreements.
And we have quite a few customers that are working with our systems, some actually using us today as part of their business operations. You're actually going to hear from NTT Docomo a little bit later about the application that they developed for optimizing cell tower resources. Actually optimizing how the paging messages flow between the tower and the cell phones, yielding better results faster, allowing them to support more phones per tower in peak periods.
North Wales Police is a very interesting customer, obviously in North Wales in the UK. Basically the problem is where to position police and other public service vehicles so that if there's an incident they can reach the incident quickly.
And their target was that 90% of all incidents need to be reached within a certain amount of time. Okay. Well, leveraging our quantum system, first of all, the time to compute the placement over the course of a year went from four months to four minutes. They can now compute placement for a year in four minutes, which also means they can reoptimize as the year goes on.
Secondly, with our solution, the average response time was cut in half. So the vehicles were getting to the incidents on average twice as fast. And finally, the number of incidents that could be covered within the allotted time not only achieved their 90% mark, but it was an increase of about 10% over what they had been getting. So really excellent results.
Japan Tobacco you'll also hear from a little bit later today, where they're focused on using quantum as a part of their AI process in drug discovery, finding that introducing the quantum processor results in more samples that are potentially more well-suited to human drugs.
I've talked a bit about different applications, primarily optimization-oriented applications, and these applications are all what I would characterize as evolutionary applications. They're evolutionary in the sense that customers are actually solving them today, it's just that they're so hard, they're exponentially hard. The compute required grows exponentially with the size of the problem, so they're trying to simplify the problem or maybe come up with heuristics to try to develop what they hope are good enough solutions. But with our quantum systems, they're actually able to solve the original problem, get better solutions, and get solutions much faster than they were getting classically.
But these are evolutionary problems because they're basically problems that are being solved by businesses today. It's just that we can do better, we can do faster, and that generates an ROI. Then there are more transformational applications, things that we think really could fundamentally transform industries. Two of them are listed on this chart. These two are listed because we are actually working on both of these today and in fact getting very close on at least the first one, the blockchain.
So I already talked about how we were able to take the computation that we did to demonstrate quantum supremacy and turn that into a hashing function that could then be used to create an entirely new blockchain architecture based on quantum proof of work. And we actually then went ahead and built this blockchain and have it running as a distributed quantum application across multiple of our quantum computers. But we really think that this could be very transformative as all blockchain today is based on computations that are susceptible to quantum attack when we get quantum computers large enough.
Now, we're not there yet and we won't be for a number of years, but eventually we will be. And so, all the cryptocurrency that's based on proof of work, all the ledger blockchain that's based on proof of work will need to transition to quantum-safe. And then when you add on top of that the fact that the power consumption for quantum is a fraction of what it is for classical, in particular GPU computation, which is what's being used today, we have the potential to really transform an industry and dramatically reduce power consumption.
We're working on exactly the same thing for AI. I mentioned that we already have made available enhancements to our software development kit to introduce our quantum systems as part of the AI machine-learning workflow. We did it for a learning model called restricted Boltzmann machines, where we are actually able to embed the RBM in our quantum system and use it to train the model to be able to generate a model that creates samples with the same distribution as the original data set.
Okay, this is really important for a couple of different reasons. First of all, much less energy consumption using the quantum computer than GPU farms. Secondly, what the quantum computer is really, really good at is very quickly generating a lot of good samples, right? So, not one at a time, but thousands of samples in one computation, right? Which also speeds up the process and further reduces power consumption. And we're now working on bringing that same capability to other AI models, including diffusion models and transformer models. And we think that this could also be quite transformative for the AI industry.
And then I mentioned that we had sold our first system to the Jülich Supercomputing Center in Germany. We also recently announced that we had signed a memorandum of understanding with Chonnam National University in South Korea, Incheon City, and ETRI to basically collaborate on important new materials and biopharma applications that also leverage the quantum computer together with supercomputers or HPC systems. And we're very much looking forward to that partnership and hopefully resulting in the first D-Wave system here in the Asia-Pacific region.
Okay, now I am thrilled to be able to introduce to you all someone who probably doesn't need all that much introduction, but Hidetoshi Nishimori, I'd like to ask you to join me on stage.
So, for those of you that are not familiar with Nishimori-san, he's a professor emeritus and specially appointed professor at the Institute of Science in Tokyo. He received his doctorate in physics from the University of Tokyo in 1982, where his thesis introduced the well-known Nishimori line in spin glass theory. Following post-doctoral research at Carnegie Mellon and Rutgers, he joined the Tokyo Institute of Technology faculty. In 1984, he rose to become a professor and dean of the School of Science and Engineering. His research has focused on statistical physics and quantum information science, making key contributions like finding exact solutions to spin glass theory and proposing quantum annealing.
He's actually the inventor of quantum annealing, which is what almost everything that I've talked about up until now is based on. So we all owe him a real debt of gratitude. He's been recognized with numerous awards, including the IBM Science Prize, the Nishina Memorial Prize, the CNRS prize, and the Medal with Purple Ribbon awarded in the name of His Majesty the Emperor. Welcome, Nishimori-san.
Thank you. So, we are basically just going to do a Q&A. The only twist is that I'm going to ask the questions in English and then Nishimori-san will translate the question to Japanese and answer it in Japanese. So, you all who are listening to Japanese translation, you can take it off for a little while. It's going to make it interesting for me because I have to figure out when he's finished answering the question. So, we'll see how this goes. But anyway, okay. So, first of all, Professor Nishimori, you and Kadowaki-san first proposed quantum annealing, and that was maybe 30 years ago. And so I'd just be interested in your perspectives on how this has all evolved over the course of the last 30 years.
Thank you very much. Please allow me to answer in Japanese, as you said.
And just so you don't think I'm clueless, Professor Nishimori did write out an English version for me so that I have some sense of how he's answering these questions. Okay. So, as you think about quantum computing going forward, what's exciting to you, or what are you looking forward to next from quantum computing?
So, how well known do you think it is that quantum annealing was actually invented here in Japan?
So I mentioned a little while ago there's some evidence that quantum annealing will always outperform gate model on optimization problems. Many believe that. Not everybody does. I'm wondering what your thoughts are on it.
And you were involved in the first demonstration of coherent quantum annealing at scale. What do you think is the significance of that work?
And then there's the recent supremacy paper that we published in Science. Do you see any areas of research opening up as a result of this work?
And what advice do you have for Japanese business leaders who might be thinking about starting to use quantum computing? Well, that's a big question.
Thank you very much.
It's amazing to think that an entire industry is being created on some of the early work that he and his colleagues did in this area. It's very exciting. Okay, just to finish up, if anybody is interested in working with D-Wave, we make it very easy. We do have a quantum cloud service that, as I said, many of our customers use to access our quantum systems. It is available here in Japan. Of course, if you want to buy a quantum computer, we will let you buy one. They'll cost you a little bit more, but we do have systems available for sale. And as I mentioned, we have a services team that is available to help understand which applications can most benefit from quantum and then how to build out those applications to get real performance leveraging our quantum systems and services.
So with that, I'll just summarize that we believe that D-Wave is the first company in the world to actually realize quantum computing. We do have applications that are being used by customers as part of their business operations today. The Ford auto-sign application that Jeremy showed you a few minutes ago is an example of that. We're the only company that actually has demonstrated quantum supremacy on a useful real-world problem. Not a contrived synthetic benchmark, but a useful real-world problem.
Most of our work today commercially is leveraging our annealing quantum computers. Those are the systems that are large enough and powerful enough to actually do useful work in the market. Ultimately, we will also have gate systems, and then we will be supporting both for the different applications that each is good at addressing. Our systems and our services are highly reliable, including offering service level agreements, and we have over a hundred customers, many of them commercial, but some of them also research and education. With that, thank you very much for your time.
