00:00:01Dr. Kristel Michielsen: We have for quite some time the vision that you have to combine the best of both worlds. So the world of quantum computing can maybe solve particular problems better or more energy efficient, but for the classical jobs and some of the hard jobs, we still need our high performance computers. They will not disappear, and therefore, our vision is to combine both.
00:00:33Murray Thom: Hello, and welcome back to Quantum Matters, where quantum computing gets real, from D-Wave. I'm your host, Murray Thom, as we move past the hype and the theoretical to explore practical real world applications of quantum computing today, and where the biggest opportunities lie in the future. Let's open the box and see what's possible.
We tend to experience computing as something invisible. We ask a question, run a search, train a model, and somewhere behind the scenes, enormous machines do the work. At the frontiers of science, those machines have a name, high performance computing or HPC. These are supercomputers built to handle complexity at a scale everyday computers can't touch, modeling the climate, new materials, biological systems, and even quantum physics itself. But the next chapter may not come from one machine replacing its predecessor. It may come from combining different kinds of machines, classical supercomputers and quantum computers, so each can do what they do best.
That's exactly the work happening at the Jülich Supercomputing Center in Germany, and it's what we're here to dig into today. Jülich is one of the world's leading HPC centers, and it's home to JUPITER, Europe's first exascale supercomputer, a system that crosses the threshold of a quintillion calculations per second. That's a billion billion calculations per second. Through JUNIQ, its unified infrastructure for quantum computing, it also gives researchers access to quantum systems, including D-Wave's Advantage annealing quantum computer.
Jülich has been a pioneer in bringing quantum into the HPC environment, working with D-Wave for years, and in 2025, it became the first HPC center in the world to own a D-Wave Advantage system. Today, as you can probably tell if you're watching, we're taking the podcast on the road, and joining me for this conversation is professor, Dr. Kristel Michielsen, director of the Jülich Supercomputing Center, head of HPC for Quantum Systems, and head of JUNIQ.
She's a computational physicist and a leader in quantum simulation and benchmarking with record-setting quantum computer simulations to her name and a framework for measuring how far quantum technology has actually come. Kristel, welcome to Quantum Matters.
00:02:40Dr. Kristel Michielsen: Thank you very much, Murray.
00:02:42Murray Thom: Now, you and I were just at International Supercomputing in Hamburg, Germany together. It was a very busy show, huge, huge systems available. I saw you giving presentations and meeting with folks. You were very busy. I was excited to have an opportunity to at least talk with you for a few minutes there. Can you give folks a bit of a sense for what supercomputers are? And if they were standing in the presence of a supercomputer, what does that feel like? Because really, you get that sense of the energy and the power of those systems when you're near them.
00:03:12Dr. Kristel Michielsen: Yes, indeed. Normally, we are used to our laptops, our personal computers, and they are rather small. But if you go to the supercomputers, you have to imagine the exascale supercomputer you have been talking about, JUPITER. This one has 24,000 GPUs, so this means that this is also a relatively large building containing all these racks of computers. And what is special about JUPITER is that you can use for one calculation all these GPUs together, if you of course manage to find a problem that is as big needing this amount of processors.
00:04:00Murray Thom: Yeah, I think that's a good point. Even a single user can set a task for those 24,000 GPUs simultaneously.
00:04:09Dr. Kristel Michielsen: Indeed.
00:04:09Murray Thom: That's the kind of unprecedented resource it represents.
00:04:12Dr. Kristel Michielsen: Yes, indeed. Yeah, because we have the hyperscalers where of course you can buy your compute time, but usually, they have divided their, we can also say supercomputers in smaller parts so that every customer can find a small part of the computer. We say in some occasions, as I said, you need special problems for this, you can use the whole machine for a certain time, because the whole machine is connected. Every processor is connected to others.
00:04:46Murray Thom: Now tell me, you're talking about these special problems that people are running on supercomputers. What are the types of applications that are being accelerated by these large systems?
00:04:57Dr. Kristel Michielsen: One example is together with a large team of researchers, we won the Gordon Bell Prize for Earth simulations, and it's with a resolution to 1.25 kilometers. So that's quite a small resolution. Everybody can assume what 1.25 kilometers means. So that's one important application in the sense that you can also use this or use the results for climate predictions and weather predictions, and on a more local scale, so that's just one example.
Now, JUPITER is a machine that it's not only therefore doing simulations. It has the NVIDIA Grace Hopper superchips, which means that it's also perfect for artificial intelligence applications, so we can train huge AI models with this machine too.
00:05:58Murray Thom: And when we think about the impact of these kinds of calculations, as you were talking about in terms of simulating the climate of the Earth to within a resolution of a kilometer, this is probably really informative for setting policy and for organizations who are then trying to use weather data when they're thinking about their supply chains. This has an impact beyond just the researchers using it, I'm thinking.
00:06:24Dr. Kristel Michielsen: Yes, you can think of applications for flooding. If you can make some predictions that there is bad weather coming already with this resolution, maybe you can also see whether particular regions in your neighborhood are in danger for a possible flooding. On the other hand, if you are talking about climate predictions and also the weather predictions, this can also have some applications for the farmers, because they know whether they have to try to protect their crops if possible, or if they have to irrigate their lands or if they can wait.
00:07:09Murray Thom: There's one other dimension I want to ask you about, which is the energy and electricity that it takes to run these large computers. So can you give folks a bit of a sense of the scale that we're talking about?
00:07:19Dr. Kristel Michielsen: Yes. So with JUPITER, we can say that this can use between 11 to 17 megawatts, but this is the energy consumption from small cities. So this is huge, and therefore, we are in Jülich also very proud that JUPITER is already, even if it's using so much energy, the most energy efficient in its class, so in its exascale class. But nevertheless, we are looking for systems that can reduce this energy consumption.
00:08:03Murray Thom: I want to now introduce quantum computing into this conversation because in 2025, D-Wave was able to publish peer-reviewed results in the Scientific Journal of Science for application test results where the D-Wave Advantage2 quantum computer outperformed one of the world's largest supercomputers, and one of the things that just amazes me is that for some hard problems, these massive supercomputers can be outperformed by a single quantum computing chip the size of our thumbnail using less than $1 worth of electricity. When you saw that result, what did you think about it?
00:08:36Dr. Kristel Michielsen: We know that this is possible. Of course, it depends on the type of quantum computer we are looking at, and the D-Wave quantum annealer is one of the bigger systems. It's also a special type of quantum computer. As I said, we know it's possible, but it's only therefore special type of problems, and we want to extend the range of applications, and therefore, we have already for quite some time the vision that you have to combine the best of both worlds. So the world of quantum computing can maybe solve particular problems better or more energy efficient, but for, let's say, the classical jobs and some of the hard jobs, we still need our high performance computers. They will not disappear, and therefore, our vision is to combine both. We have to integrate them, and then we can give part of the problems to the quantum computer, which it can solve most efficient, maybe faster, maybe more energy efficient, or even better, and then we use the rest of the HPC system for solving the rest of the problem. Then we have solved it in the most efficient way.
00:09:55Murray Thom: Yeah. Yeah, and this is a conversation I'm frequently having on the floor at International Supercomputing is that quantum computing is not trying to target the applications that run efficiently on classical supercomputers. It would never be able to catch up just from a technology development standpoint. So it's an opportunity for us to take problems which aren't well-suited, let's say, to floating point operations, and bring a new resource which maybe is handling them better, and as you're saying, really then as a result, increasing the energy efficiency of the whole application.
00:10:26Dr. Kristel Michielsen: Yes, that's true. And in our vision, we always say a quantum computer is not a universal computer. Maybe it can do any calculation, but the point is never add numbers with a quantum computer, at least with the quantum computers we have nowadays. This is still something which a classical computer should do, and if you have a lot of these additions, then use a high-performance computer.
00:10:56Murray Thom: Well, I think that's a very good point because I think that can often get confused in the quantum computing space, because there's one question which is could I run the problem on a quantum computer? And then the second question is can I accelerate that application with the quantum computer? And universal is really the first question. It doesn't have to do with whether it can be accelerated, and so that concept of general purpose computing, quantum computers are not going to get used for running spreadsheets and sending email, but to some extent, they're carving out a portion at the frontiers of computer science an application space, a set of things that they're exceedingly good at, and that defines their domain of acceleration. And it separates, I think, and makes more clear what the domain for classical computing acceleration is.
00:11:42Dr. Kristel Michielsen: Yes. And that's actually also why researchers have been thinking about quantum computing, because we know we have our huge classical systems, our high-performance computers, but we still have problems which we cannot solve with them, or it takes too much time or it requires too much resources, too big systems. And examples there are quantum systems, and quantum systems we find in quantum chemistry, we find them in physics, we find them in a lot of scientific domains, and we know we need a more efficient computer to calculate or to do computations for these research fields. And then the idea was a quantum computer is a quantum system itself, so maybe this is the optimal computer to solve such type of problems. That's how we came to quantum computing.
00:12:43Murray Thom: The ideas you're sharing here about simulating quantum physics with computers is actually touching on the original motivation that Richard Feynman proposed, which is that if we want to be able to do some of these calculations, the inefficiency is so high, when we try to break it down into floating point operations, we really need to build computers that can exhibit quantum properties themselves.
00:13:03Dr. Kristel Michielsen: And then there was one thing in addition, and that's we have quantum theory, and in quantum theory, we can prove with pen and paper that we have something like quantum parallelism, which could accelerate the classical computation. But then the question to the engineers is can you now construct a device which operates in a way as what we have designed mathematically with pen and paper? And that's of course a challenge.
00:13:37Murray Thom: Yeah. Yeah, exactly. The way I tried to make this clear is that we have this amazing behavior in the world that has been hidden to us for most of the entire history of the world, and when we started to be able to control and manipulate things at these tiny levels, we saw this behavior and we needed to develop entirely new mathematics to describe it. When we think about when I'm describing a tree that I see outside, I can use the English language to describe the tree, but when I'm trying to describe the behavior of quantum mechanical systems, I actually need mathematics to be able to fully describe what its capabilities are. And we can now use that mathematics to actually design new computers so that we can turn those quantum effects into useful work, but we're still struggling to understand what the implications of that mathematics are, what's actually taking place in the world.
00:14:27Dr. Kristel Michielsen: Yes. And actually, you know Murray, the issue here is when in the past, we could build or the engineers could build a steam engine and it was working, but nobody understood how it was working, so the theory came afterwards. Now here, we have a theory, and we say according to this theory, we can do something wonderful. We can compute and maybe much more efficiently as we can do with computers that we know nowadays. But then the question is, engineers, can you now build according to this theory? And of course, this was also something which they had never done.
00:15:11Murray Thom: I'm really happy that you made that connection to the origin of the steam engine, because I'm often sharing with people that a quantum computer is humanity's creation of a new kind of engine. It's an engine that's turning quantum mechanics into useful work, and we're now making something that used to be very difficult for us to do extremely easy and energy efficient to do, and there are probably a variety of applications we haven't even discovered yet because we've never had this capability.
So given this point in time, I think it's probably important to touch on the strategy, the quantum computing and classical computing strategy at the Jülich Supercomputing Center that you are forging and creating, because as I'm traveling around the world, I often share that Jülich vision for how to learn about quantum computing and what kind of investigations to do, because it's a great example for other people. Can you talk to us a little bit about that strategy?
00:16:03Dr. Kristel Michielsen: We have been working on quantum computing for quite a while and simulating quantum systems on the high-performance computers. But then when the technology evolved and one can really do something with it, we thought then we need a strategy, and our strategy was we keep on modeling these quantum computers, so meaning try to model what they are doing into some mathematical language so that we can then take this and put it into an algorithm which we can run on the high performance computer. That's one thing.
Then our second component was we want to host quantum computers because we want to be very close to these systems. We have been hosting our HPC systems for such a long time. These are new systems. It's not only important for the ones who are using the systems, but also for our infrastructure people. They need to know what does it mean if you have helium cooling? What does it mean if you cannot have vibrations because it's disturbing a quantum computer and so on. So that's then another, let's say, pillar of our strategy.
And then we know from classical computing, every time you have a new type of processor, for example, if you go from CPUs to the GPUs, then many people had their algorithms, their software written for CPUs, and all of a sudden, there were new processors, so they had to adapt their software. As a computer center, we have put a lot of efforts in helping our users to transform their software codes to the new type of computers, and we see the same for quantum computing. We need to help our users in translating their problems and really helping them to use these new type of computers.
And for us, of course, we want to continue doing research, also develop applications ourselves. So then we said, "Let's set up a user facility for quantum computing and not focus only on one type of technology. We want to make our users familiar with as much technologies as possible." They need to see what the different machines are doing, what the advantages and the disadvantages are.
00:18:48Murray Thom: Well, and I think from just speaking from my experience, what's so important is that this strategy lays out a roadmap for folks about how to break into an entirely new information processing category, because it's about understanding the machines, it's about identifying the users, giving them access to the systems, helping them translate their codes, and then understanding how that impacts the domain of applications. I want to touch on this point associated with users here, because certainly when I'm talking with supercomputing centers, they have a user base that have been working in high-performance computing for a long time and they're wondering, what is the domain of users for quantum computing systems? What's your answer to that question?
00:19:31Dr. Kristel Michielsen: That's a very good question. So we have a lot of users in HPC, in the meantime, also now in AI modeling, building AI models, using them, training them. But for quantum computing, the issue is these users might benefit from quantum computers, but maybe they do not know yet. Maybe it's not yet for production. So you really have to inform these users to challenge them and say, "Maybe this is also something useful for you. Maybe not now, but maybe for the future, but it's now that you need to get familiar with it. You still have some time which might be worthwhile to invest right now."
And we challenge also our users to look into their HPC workflow. So how is the program organized? And is there maybe a small piece which you can give to a quantum computer and calculate it in the hybrid mode? So use the quantum computer for this tiny piece and then let the HPC system do the rest of the work. And then they can stay in their comfort zone and only touched it, the quantum computer, for a small piece.
00:21:02Murray Thom: There's levels that I'm hearing in that answer as well, Kristel, because certainly at D-Wave, and I think also in that story you're telling about at Jülich, you've got users who are focused on their application, and that's where they should be. And really, then it's about by having a team of folks at Jülich who understand the technology, they bring the appropriate technology to that application for those end users so that they're not having to start with the technology and then work towards the application. And then I think also, there is a lot of, let's say, folks who understand a lot of low level details about the architectures of classical computers who are focused on the most efficient way to break problems down into floating point operations. There's now an opportunity for a new community, the community that you're building to say, "Hey, actually, we're not going to be breaking some of these things down into floating point operations. We're going to fit them into quantum machine instructions." And so I think that's an opportunity to bring new users to the space.
00:21:58Dr. Kristel Michielsen: Yes, definitely, and we also have seen in-house that these things can work perfectly. So for example, we have here some biologists, of course, computer biologists who did not have any experience with quantum computing, and they were interested because we have given several presentations, trainings and so on, and they were then interested in doing some protein folding on the D-Wave quantum annealer. And they asked, "Is this possible?" And we said, "We didn't do it, but if you are interested, we can help you. We can assist you in translating these problems, but you might have to accept that you have to adapt your problem to the hardware." And that's actually what they did.
00:22:58Murray Thom: You've touched on this before here, so the integration of quantum computing. I think that the question of how to bring high-performance computing with quantum and also AI is a topic on a lot of people's minds. What does the road ahead of us look like from your perspective?
00:23:13Dr. Kristel Michielsen: Yes. So our plans is also to do a tight integration of the D-Wave quantum annealer with our HPC infrastructure, as I said, because it's a reliable system which is running 24 hours a day, seven days a week, so in that sense, you can also connect it to a larger HPC system. But we want to do this because D-Wave anyway has the hybrid mode of simulating, but we want to bring it to a very tight integration. But there are still challenges, and challenges are how to do the job scheduling. Because usually in HPC systems, we are working with the queues in order to give all users their share of the machine for a certain type. Now, we have to see if you then use different compute technologies, the HPC system and the D-Wave quantum annealer, we don't want the HPC system to wait until the D-Wave annealer is finished, but also not the other way around.
These resources are too valuable not to do anything and are in waiting mode. That's what one wants to avoid. So there are challenges, and at the beginning, one can accept waiting times if these are not too long, but in the end, it has to go also in a production mode.
00:24:52Murray Thom: That's true. That's true. The history of that integration hasn't taken place yet, so it's those frontier open questions we get to answer by looking at that. Now, you're touching on this. At Jülich, at your center, you've hosted and operated supercomputing systems for decades, and you've now had the opportunity to operate quantum computing systems for years as well. What has it been like hosting a quantum computer relative to the other machines in the center?
00:25:16Dr. Kristel Michielsen: This depends on the type of quantum computer. So as I said, the D-Wave quantum annealer is a machine with which we have in the meantime long-term experience. As it's running 24 hours a day, seven days a week, it does not require a lot of maintenance, so this means it makes it easier to make this integration, because in that sense, it's a little bit similar to the high-performance computers that we are hosting.
For other types of quantum computers that I would say did not get to reach this level of operation in the sense that it does not have this high level of availability during the day because it needs recalibrating or it even stops working for a couple of days, that makes it a little bit more difficult to integrate. One can integrate, but then one has to take into account that the hybrid modes might be off for quite some time because the quantum computer is failing. So that's our experience, but as I said from the beginning, the D-Wave quantum annealer is there for quite some time. It has been scaling over the years. Many of the quantum computers, if we don't talk about the biggest companies with the biggest gate-based systems, are in a completely different position. We are also talking about much smaller systems.
Is it valuable to integrate them? Yes. In our opinion, yes, because we still do not know which quantum computer will be in the race, which will be the most efficient one, and integrating them brings another dimension to this. It might be that the quantum computer itself might not have the best qualities, but as an integrated system, one does not know. It might even, from an energy point, then still be better than another one. As I said, they are not universal, so we see them as special purpose machines. And maybe then an ion trap technology or a superconducting technology might behave differently in this hybrid environment.
00:28:00Murray Thom: Yeah, so many open questions to be developed. Now, I have a question for you, which is maybe a difficult question to answer. You'd be the judge here, but I'm thinking that quantum computing is a new entrant into high-performance supercomputing in terms of the calculations it's able to do in magnetic materials and quantum simulations. I'm wondering, are there lessons that the high-performance computing community has learned that would be good lessons for the quantum computing community through its development? Are there transitions and milestones that high-performance computing on the classical side went through that maybe the quantum computing community is going to see coming in its future?
00:28:41Dr. Kristel Michielsen: I would say for the HPC community, it has been a question of scaling. At the beginning, we had computers, and actually, then there was more or less the same question. We have a computer. What to do with it? AI was a similar thing. So at the very beginning, AI was also very limited. We here at the Jülich Supercomputing Center have hired some people and their task was bring AI to the big HPC systems, so use it on a large scale. And nowadays with this AI hype, people think this has always been like this, but it has not always been like this. We have experienced this transition here at the computer center in AI, and it happened before also with HPC. Maybe the same will happen with quantum computing.
We have there now a similar question. Scaling one chip in a quantum computer might not be possible. Linking several to one big quantum system might become possible. We know that companies are working on this. The challenge is it has to be one quantum system, but if that works, we have a scalable quantum computer, and maybe then we are also working with quantum computers in a different way. So there are similarities. Yeah, there are similarities.
00:30:17Murray Thom: Kristel, it's been great having you on the podcast and getting a chance to share our mutual excitement for quantum technology. Thanks for being a part of Quantum Matters.
00:30:24Dr. Kristel Michielsen: Yeah, thank you, Murray, for hosting me. It was a very interesting discussion. Thanks a lot.
00:30:33Murray Thom: What a pleasure and an honor it was to have Kristel on the show today, for her to be able to share from her experience with high-performance computers and quantum computers, and the fact that they solve different types of problems. So it's really not about quantum computing replacing supercomputing. It's about combining high-performance computers with quantum computers so that they're each contributing their strength in a way that helps accelerate applications and make them more energy efficient.
The other thing I think is really important is how they're engaging users, giving them opportunities to have experience programming quantum computers alongside supercomputers, and also those users who are thinking about it from an application level, understanding how they can see that acceleration so they can get to their objectives and goals with the applications that they want to run.
And then I think the other key thing that Kristel shared is that supercomputing started really CPU-centric, focused around a particular type of technology, and it has expanded to include GPUs and quantum computers and neuromorphic computing, where Kristel was able to share with us that the future is really about heterogeneous computing, so that's where really advanced computing, the frontiers of advanced computing are headed.
That's it for this episode of Quantum Matters. Thanks to you, our viewers and listeners for joining me. Please follow so you don't miss an episode, and to learn more about how D-Wave works with organizations to get started and succeed with quantum, visit dwavequantum.com. Until next time, I'm Murray Thom. Stay curious about your quantum reality.
