Notice: We are in the process of migrating Oral History Interview metadata to this new version of our website.
During this migration, the following fields associated with interviews may be incomplete: Institutions, Additional Persons, and Subjects. Our Browse Subjects feature is also affected by this migration.
We encourage researchers to utilize the full-text search on this page to navigate our oral histories or to use our catalog to locate oral history interviews by keyword.
Please contact [email protected] with any feedback.
Photo courtesy of Microsoft
This transcript may not be quoted, reproduced or redistributed in whole or in part by any means except with the written permission of the American Institute of Physics.
This transcript is based on a tape-recorded interview deposited at the Center for History of Physics of the American Institute of Physics. The AIP's interviews have generally been transcribed from tape, edited by the interviewer for clarity, and then further edited by the interviewee. If this interview is important to you, you should consult earlier versions of the transcript or listen to the original tape. For many interviews, the AIP retains substantial files with further information about the interviewee and the interview itself. Please contact us for information about accessing these materials.
Please bear in mind that: 1) This material is a transcript of the spoken word rather than a literary product; 2) An interview must be read with the awareness that different people's memories about an event will often differ, and that memories can change with time for many reasons including subsequent experiences, interactions with others, and one's feelings about an event. Disclaimer: This transcript was scanned from a typescript, introducing occasional spelling errors. The original typescript is available.
In footnotes or endnotes please cite AIP interviews like this:
Interview of Krysta Svore by Will Thomas on September 17, 2024,
Niels Bohr Library & Archives, American Institute of Physics,
College Park, MD USA,
www.aip.org/history-programs/niels-bohr-library/oral-histories/48491
For multiple citations, "AIP" is the preferred abbreviation for the location.
In this interview with Krysta Svore, Technical Fellow in Microsoft’s Advanced Quantum Development Team, she reflects on the development of her work on quantum algorithms and leadership role in Microsoft’s quantum computing efforts. She discusses her undergraduate education in mathematics at Princeton University, doctoral education in computer science at Columbia University, postdoctoral work at MIT, and an internship at IBM. She recalls working with figures such as Isaac Chuang, John Preskill, David DiVincenzo, and Barbara Terhal, and gaining an understanding of the practicalities of constructing quantum computers. Svore explains the origins of Microsoft’s efforts in quantum computing with Michael Freedman and Staton Q, and she discusses her own efforts to develop software in tandem with the company’s efforts to develop quantum computing hardware. She reviews the effort to integrate quantum computing into Microsoft’s Azure cloud platform, the roles of the Q# programming language and the company’s partnerships, and how Microsoft benchmarks quantum technology. The interview concludes with her reflections on the status of women in the field of quantum computing.
This is Will Thomas with the American Institute of Physics. The date is September 17, 2024. I am online with Krysta Svore of Microsoft, and we are going to be talking about her experiences in the field of quantum computing, and also, through your experiences, learning a little bit about the development of the field as a whole over, say, the last 25 years. So, to begin with, could you, for the record, say what your current title is?
I’m Technical Fellow in the Advanced Quantum Development Team of Azure Quantum at Microsoft.
And you are at least recently a vice president there as well, is that right?
Yes, yes. Also, Vice President of Advanced Quantum Development.
Okay, great, so you still have that.
Yeah. Technical Fellow is actually a better thing than VP. [laughs]
Well, we’ll get to your ascent through the ranks in a little bit, but let’s take it back to school and how you got interested in science and mathematics, how you may have originally heard about quantum phenomena, all that sort of thing.
Yeah, so I would say I’ve always been fascinated by puzzles, by challenging things in science. I would say even in high school, I was always interested in research and then how that research applies to people and humanity. So, in high school, I interned at Fred Hutchinson Cancer Research Center and studied and wrote my first research paper with an amazing woman there, Jan Chalupny, so studied more about cancer research at that time, thinking I would go into molecular biology and that type of research for my career. But when I entered college, I had taken very limited computer science in high school, but I majored in mathematics, and I also started studying computer science as what would be the equivalent as a minor. And as I learned more about computers, I learned you can do research with machines, right, with computers. You don’t have to be in a wet lab or working with, say, animals or actual beings. You can do it mathematically in a machine. And so, I loved this idea of doing research with computers, with computing. And the idea that computing could enable you to do things you can’t just do in the lab. It can help you go beyond and find new discoveries. So, I became fascinated by this, and at the same time, I was studying mathematics as my major.
So, just to put a roadmap on this, where did you grow up? And I know you went to Princeton, correct?
I went to Princeton for my undergraduate. I grew up in the Seattle area. So, I grew up with Microsoft in my backyard, so I knew Microsoft growing up. It was a younger company than what it is now, much smaller campus. But I grew up knowing about Microsoft, of course. But I went to Princeton as an undergraduate, and I had a junior seminar there from Professor Andrew Wiles, who is now known for solving Fermat’s last theorem. And Professor Wiles had us do a seminar on cryptography. And it came up in one of the lectures that there was this model of computation, this type of computation called quantum computation that only a couple years earlier had been shown had the ability to break what we know as our mainstay of e-commerce today. It had the ability to break RSA.
And I was amazed that there was this other type of computing, not the kind of computing we were using, digital computing, what you might call classical computing, that we were using daily. There’s this other model that relied on quantum mechanics and the attributes and characteristics of quantum mechanics to bring about new properties, right? New types of operations, new ways of encoding information, new ways of storing that information and operating on it. So, namely, quantum information, right? And a quantum computer could operate on this information and give you access to solutions that you couldn’t have access to in any efficient amount of time with classical computers. So, I thought this was incredible.
This was around the year 2000 we’re talking about?
Yeah, I think I took this class in 1998, 1999, something there. So, right around 2000. And Peter Shor’s algorithm, called Shor’s algorithm, as we call it now, was this algorithm, this algorithm that used quantum computing to show that you could do this factorization problem, which underlies RSA. And so, that was just a couple years earlier. So, people were still digging into this paper, trying to understand how it worked. And so, it was kind of this convergence of learning about mathematics, learning about computer science, and then also learning about the existence of this other model of computation. These machines weren’t built yet, these quantum computers. But all three of these things came together for me, and I thought, “Wow, how cool would it be to figure out how to program”—first of all, someone had to build the machine, of course. But gosh, could you program it to do this? How would you communicate with this new type of machine? How would we program it? How would we map to it, compile to it, and have it actually compute such an algorithm?
So, I really liked the idea of thinking about programming languages for this machine, the software stack for this machine, and error correction for this machine, this yet-to-be-built and architected machine, right? And there were people working in physics departments working on qubits. How would you create a quantum bit? How would you create information that could be stored in a quantum bit? And these quantum states. So, there were folks working on that at a variety of places at that time, but nothing like where we are now, where you actually read about quantum computing, say, in your mainstream newspaper, right? It was very much this very small community of folks thinking about this. And so, I feel I was really lucky to have learned about this at that time and had the opportunity to really start to study it before it was big, right? Before it was mainstream.
Yeah, you were looking way out ahead of some of these problems.
Yeah, and some people have been studying it for a number of years already. Of course, amazing people in the field today. But it hadn’t really reached the level of, I would say, computing yet or computer science, if you will. Like, people in computer science looked at it, but none of these stacks existed yet. And we didn’t have the level of classical computing we have now, so there weren’t really a lot of simulations, or emulations, or other things one could use to study this. So, it was a wonderful time to enter the field and have, really, kind of a whole field of opportunity, of low-hanging fruit, if you will.
So, you then went on straight to Columbia University, is that right?
Yeah, so of course, I became so fascinated by this, [laughs] I decided I would make it my focus. And the best way to make it your focus, in this case, really felt like it should be to go get a PhD and focus a dissertation topic around this, formulate a thesis on what to do. And so, my dissertation was on fault-tolerant, scalable quantum computing, how to achieve it. Still working on that today, right? Went to Columbia, but I had this wonderful fellowship, and this fellowship enabled me also to spend large amounts of my time…
What was the fellowship?
It was through the National Physical Science Consortium, and it was for women and minorities. So, I received a fellowship that essentially covered my PhD. So, I was able, then, to move a little bit during my PhD. So, I received my PhD from Columbia, under Al Aho and Joseph Traub, and at the same time, I was able to go to MIT for, I think, almost a year, a number of months there, working with Isaac Chuang in his lab. So, that was an ion trap lab, so I got to go, and sit, and be a part of hardware development, thinking about software for that hardware, thinking about how you error correct that hardware and that noise. I was able to also spend time at Caltech for a few months with John Preskill and, again, learn more about quantum error correction and quantum information, work with an amazing group of people there. And so, again, I just felt so lucky to have a PhD in which I could actually really explore and work with a number of the field’s experts, right? So, again, like John Preskill, Isaac Chuang.
I also spent internships at IBM, working, at the time, with David DiVincenzo and Barbara Terhal, so these people were just amazing to be able to learn from them. They’d been in the field for a number of years, looking at this, and just had so many wonderful perspectives to learn from. And then, after that, I ended up joining Microsoft research, and quantum computing was in a very different place than it is now. I studied error correction, right? My thesis really focused on error correction. How would you combat noise, right? Noise is one of the biggest challenges still in quantum computers. And so, looking at how you combat that noise, how do you write software and communicate with the machine such that you can really minimize the noise as well?
I’m curious about the nature of the work. It’s obviously primarily theoretical. Does it tend to be on the physical side of things, mathematical, logical, some combination?
For my work, I think of myself as, you might say, a bit applied. By training, I’m a theorist, but I definitely migrate towards the application side. I want to see theory put to practice. I want to see it made real. That’s my passion, seeing the machines light up, seeing users be able to use them and drive new solutions, new discoveries from them, right? And so, no matter the area I’m looking at, because I also worked in the space of machine learning very early in Microsoft for a number of years, there, again, I’m fascinated by disruptive technologies, but making those disruptive technologies very usable so that they’re bringing solutions to humanity, to our world, but really bringing them forward, right? Not just staying in a box. So, the type of work is, really, for me, leveraging classical computing, actually, to enable faster studies, more studies, again, enable programming to these machines. So, how do we leverage the tools we have to bring this capability, this disruptive computing forward?
Your PhD was in computer science. This is an area where there are a lot of physicists involved. How many of you were there on the computer science side, and did you notice a strong stylistic difference between you and the people more on the physics side of this area?
Yeah, absolutely. I think as a computer scientist—mathematicians, engineers, computer scientists, physicists, at that time, I would say, first, it feels like the field of quantum computing perhaps emerged mostly from a mixture of maybe physics and math, you would say. And then, it started to broaden. You started to have a few more folks maybe on the computer science side, and then a few more on the engineering. And now, we have people from all disciplines in the field, which is necessary. But it started tighter around maybe physics and math. And I think as you point out, it’s very much the case that if you’re trained in one of these disciplines, you have a different set of maybe frameworks that you leverage, a different type of computational thinking. I think in computer science, you really think in this computational way.
In some ways, you think about how you structure a program to do the task, so you think in loops and blocks. And that’s different than perhaps how a person trained directly or mainly in physics would think. And I think that’s where, again, I think having worked with a number of folks early who were trained in physics, or trained in math, or trained in computer science, was incredibly helpful because you could learn the language, you could learn how to communicate. And so, as I mentioned, I went and worked in Ike Chuang’s lab. That was the first time I was in a lab of physicists, right? [laughs] And even that lab at the time was in the Media Lab at MIT, so it maybe wasn’t even your traditional-traditional physics. But there, you get to learn to communicate.
And he had come there from IBM already, right? He had done the factoring of 15 experiment?
That’s right, and the NMR setup very early. So, I think the opportunity to work in a lab where it was experimental was huge, right? It’s very real when you’re sitting in a lab, and you can actually see the ions being held in place. It’s real, right? These machines can be real. And of course, that was, what, 20 years ago that I was there. It was the very early days of qubits and quantum computing in actually being built.
As somebody coming into the field, did you get that sense of the freshness and openness, of it being kind of a nascent field?
Oh, yeah. Oh, absolutely. The early conferences—I didn’t go to the earliest ones. I’ve been in the field a while, but not the longest by any means. [laughs] There’s a lot of wonderful people that have studied this much longer than I have. But those early conferences, those early workshops, it was a handful of people. Now, it’s, like, upwards of 1,000 will go to the QIP, Quantum Information Processing, which is the big theory conference. But at the time, it was so small, we could hold it in this one corner—I remember one corner of the building at MIT, right? It was this, you might say, tiny room, in some sense. So, it was much, much smaller. Very intimate, in that sense. You really had the opportunity to talk to people from different backgrounds who were studying different aspects of the problem. It was wonderful. A lot of open communication and collaboration.
Can you tell me a little bit about your experience at IBM? We’ll, of course, get to Microsoft very soon. I talked to Jay Gambetta last week during the Quantum World Congress. He didn’t arrive [at IBM] until later, so I’m kind of interested in the period where you were there, what it was like.
Yeah, I had met David DiVincenzo, gosh, it was, like, a coffee, or a dinner, or something. And I just asked, I said, “Gosh, do you ever have internships? I’d love to work with you. It’s just a short train trip from Columbia, Yorktown Heights.” And he said, “Well, let me get back to you.” And what do you know, I think it was maybe a couple weeks later, he said, “Yeah, why don’t you come as an intern?” And that actually led to almost two years straight, right? I actually ended up doing quite a bit. I would go every week, for the whole year, it ended up, to IBM, for about two years when I was doing my PhD at Columbia. And so, again, Barbara Terhal and David DiVincenzo were there. Those were the folks I was working with. But many people were there, John Smolin, Charlie Bennett, just some incredible names in the field. And it was an opportunity—in working with Barbara and David, what we were able to do is think about error correction and start to think about it in terms of an architecture.
It was the first time we started to write down what would error correction look like for a real architecture that has constraints, right? Computers are never perfect. [laughs] Especially quantum computers. So, they have noise. Not necessarily can every qubit talk to every other qubit, meaning you’re not necessarily allowed to have every qubit just operate with another qubit in the system, right? Sometimes, they’re constrained. So, we actually studied, for the first time, over those two years, realistic architectures. And what would these theorized quantum error correction codes… so, folks like Peter Shor had invented a number of quantum error correction codes—Daniel Gottesman, Manny Knill, other folks, had invented some of these codes. They had come up with a theory of quantum error correction, which was fantastic, right? Error correction, they had shown, in theory, works for quantum information.
But no one had yet laid that down on a real architecture where you have constraints, right? You actually have wires. [laughs] You have communication over those wires, so you’re constrained to the layout on that chip. Much like we think about classical computers, you have some chip design, a chip architecture. You have to think about that architecture and then see if that error correction works on that architecture. And so, for the first time, we wrote, I think, kind of the first two papers here on understanding, “What if you only have a nearest-neighbor architecture?” So, think of qubits sitting on the grid, like a Manhattan grid, right? And they can only talk to their four nearest neighbors. Or, maybe they can only talk in a local region to other qubits.
Then, what does that do? Do you still have the ability to use a quantum error correcting code, or do those codes all of a sudden no longer work? And so, we showed that you do, indeed, still have what we call a threshold in quantum error correction when you’re constrained to a nearest-neighbor architecture or to a local neighborhood architecture. So, it was a wonderful time of, I would say, invention and starting to look at things—again, driving from theory to practice. Like, starting to drive it towards the more practical, taking this beautiful theory and starting to understand, what does it mean in practice? And that led to other work, too, that I didn’t necessarily do at IBM, but then later, take Shor’s algorithm, take the algorithm for RSA. Well, how many qubits does it actually require?
So, we wrote some very early papers, really studied very early on, “Hey, here’s an algorithm that’s written down, but this algorithm’s not written down in terms of what the architecture actually uses. It’s written down from the high level, what’s required at the algorithm level, but if we actually map it down to what some nice hardware architecture might require, then what will it need? So, how much error correction does it need? How many qubits?
So, we did very early resource estimates on, for example, Shor’s algorithm and a number of other algorithms similarly. And around the same time, we also, with a number of my collaborators, invented the first way to do addition on a quantum computer. So, we developed a log-depth adder. None of this existed, right? In computer science, the Knuth volumes, it’s how you do your logic, how you do your operations in a classical computer. There was an opportunity to take, essentially, the Knuth volumes and come up with how you do it in a quantum machine. And so, we did that as well. Again, this is maybe 20 years ago.
Then, it was in 2006 that you went to Microsoft. Were you certain that you wanted to go into a company-type environment, or did you think you might stay in academia? What was your decision-making like around that time?
I was very passionate about research and very convicted that I wanted to do that at an institute, at an industrial setting, at a lab, where you’re surrounded by other people studying that topic. I wanted to go somewhere where a number of people were working in this area, where we could collaborate, bring together each other’s strengths, and really advance the field. I felt, for me, the best place for that was indeed at either an industrial research lab or at an institute. So, I was looking at places like IBM, for example, at Yorktown Heights, IBM Watson. I was looking at Los Alamos National Lab, for example, I was looking at Google Research, I was looking at Microsoft Research. So, a number of places like that. And I chose Microsoft Research.
Microsoft Research had, and I believe still does, really the most computer science PhDs in one place in the world. [laughs] It was this amazing opportunity to go work with wonderful people on problems. But the other key thing for me, why did I choose Microsoft? I really wanted to make sure that what I worked on became something that people could use, and people could touch, and people could experience, and use for what they needed to go do. So, it was very much the place for me because it was this mixture of research institute, research lab, where you could do the research and the development, and you had the ability to drive that to be consumable, to actually empower people. And that was important to me. I needed to know that what I worked on had the opportunity to touch people’s lives.
Can you give me a little bit of background on the history of Microsoft’s involvement with quantum computing? You were coming into an existing group. How established was it at that point?
It’s really interesting because I joined in 2006, and I did not interview at Microsoft in quantum. I actually had done a few papers, maybe one paper, in machine learning when I was at Columbia. And to be honest, I’d studied quantum computing for my PhD. I really wanted to see it made real. And I was not sure at that time if it could be made real—full transparency, right? Because the types of qubits and the types of machines being built, they felt pretty far away from scaling to be able to do a full algorithm. And so, I wanted to make sure I could work on something that touched people, again, that really impacted people’s lives. And at the time, I had just learned about PageRank. And PageRank was an algorithm that’s now… now, AI has really ballooned and grown, but at the time, I’d learned about PageRank by Page and Brin. That’s the foundation of Google, essentially, right? And so, this algorithm looked super interesting. I’d written this paper on machine learning, so I actually interviewed at Microsoft in machine learning.
And I thought, “Okay, this is something that uses statistical physics, it uses a lot of the mathematics and the models that I’d learned, similar concepts, a lot of computational thinking, to get machine-learning models to work.” And I thought, “Okay, this is a way I could also impact folks.” And so, I ended up joining a machine learning group at Microsoft Research. And that’s because Microsoft, at the time, I didn’t know they had any quantum investment. So, I joined, and within three or four months, I saw there was going to be a talk by Michael Freedman on topological quantum computing. And my passion for quantum computing never was dark or silent, right? It was there the whole time. So, I went to this talk, and I immediately spoke with Mike after, and I said, “Oh my goodness, this is amazing. Microsoft’s investing in this.” They had just started that work, right?
Now, he was down in Santa Barbara, is that right?
He was originally in Redmond and then in Santa Barbara. Because of the physics institute there, Santa Barbara was a great location to set up Station Q. And so, Station Q was just starting. They were just setting up shop in Santa Barbara. He was up giving this talk. And there had been a small handful of folks that had studied quantum computing in Microsoft Research, but now it was really forming into a lab, a group. And so, I met Mike, and afterwards, he invited me down to Santa Barbara to give a talk. And it was kind of one of these lucky moments, right? He gave this talk, and it turned out there was this huge snowstorm. We don’t get huge ones in Seattle that often. [laughs] So, we were stuck on campus, so I actually had the opportunity to chat with Mike more about the quantum work they were starting in Santa Barbara.
And so, eventually, I went and gave a talk on quantum error correction, on these studies that I had done during my PhD. And this really led to continuing to be a part of staying in the loop, if you will, on the work of quantum at Microsoft. I ended up then getting an intern to focus on quantum computing, and one thing kind of led to another. Pretty soon then, around 2010, 2011, put together a proposal with Mike Freedman and Burton Smith, looking at, “How do we actually expand beyond theoretical physics and the theoretical work around topological quantum computing?” The focus originally of Station Q was on topological quantum computing, but very much with theory and physics in mind. And so, we expanded that to say, “What do we do with this machine? What are the applications you need? What is the software environment? What better place to do software than Microsoft?”
Was their work at Station Q purely theoretical, or was there an experimental component to it as well?
There was. In the early period of Station Q, it was really focused on theory, and there was an experimental component to it with university investments and work with universities, so a lot of collaboration with universities to do the experimental side in the early days of the program. And that was a great way to really explore a number of options, a number of possible paths, on how to bring a topological qubit forward in a device. And so, lots of collaboration with academic institutions at that time. And then, over time, as the path became, one might say clearer, or there was a very clear device path, that then shifted, and over time, we’ve shifted much of that work to be directly at Microsoft, where we hired a full experimental group of physicists to drive that work forward. And then, not just physicists, right? When you’re building a real machine, this extends far beyond just the experimental physics. It’s also materials scientists, it’s engineers, device physicists, measurement specialists. You have cryogenic specialists. A bunch of folks now all working together to build the machine.
Coming back to the proposal that you were discussing, was that for a software component?
Yeah, so again, building a quantum computer is not just the qubits. There’s a lot that sits around the qubits. The qubits are a foundation, right? Just like if you’re going to build a house, you need a foundation on which to build. That foundation is really that qubit system. And then, up from that, you need everything around it. You need to know how you control and measure the qubits, right? You need to get information in and out of the qubits. You need a system for, again, enabling users to program that machine. So, you need an entire software stack with programming languages, compilers, assembly-like languages, that entire stack. Then, around that, you need to be able to integrate with other computing capabilities. And this is really important, that you have a hybrid capability, so you have to be able to integrate with cloud computing, with high-performance computing, you have to integrate with AI models now in particular. And so, really, what we put together was a more thorough and complete system design that really showcased, “Hey, we have to understand what we’re going to do with this machine in order to design the machine well. It’s good to have an idea of the applications you’re going to run. What do those workloads look like?”
Is this getting out ahead of the technology, or is this actually informing the development of the hardware?
Both, right? It’s a very tight feedback loop. You want to get out ahead. You want to see where you’re headed. What’s going to be required there? Looking at the applications, we started to understand that really, the application space for quantum computers, and we studied many, many quantum algorithms, the sweet spot is chemistry and materials science, right? At the core of a quantum computer, you have quantum mechanics. Quantum mechanics is really great at modeling strongly correlated electrons, exotic properties of materials. And so, you want to use the quantum computer and really bring it together with what it’s good at, and that’s simulation of these quantum systems. and that’s where it really also makes sense from a resource perspective, right? So, the quantum computers required for chemistry and materials science are reasonable in scale. You’re looking at thousands of really good logical qubits there to solve these problems. And so, we really started to focus our designs, architect for solving chemistry and materials science. Now, that’s not to say you can’t go use the machine for something else. It is a universal, general-purpose quantum machine that’s, again, integrated with HPC and AI. But the architecture we’ve brought forward and how we’ve been able to co-design the system with the applications in mind, that’s been critical in reducing the resources required for a scaled quantum computer.
Do you have a formal road-mapping process? I asked this of a couple of other people last week, so I’m kind of curious about the different approaches that different companies have taken.
Yeah, we definitely set roadmaps and milestones internally for our system. That extends across that whole system, so we do have a formal road-mapping process internally.
Now, could you tell me a little bit about the development of your own career? How soon was it before you were in charge of a lot of it? Were you a member of the team, and then you became in charge? Were you in charge all along?
Again, in 2010, 2011, we expanded the quantum program at Microsoft to be inclusive of software and applications. So, at that time, essentially what emerged was, I led a team that we called Quantum Architectures and Computation, which was really part of Station Q, but then also the Station Q focus around theoretical physics and experimental physics, building the devices. And so, we had work starting to come together, and then we also had work on the control systems and kind of the infrastructure around that as well that, at the time, Burton Smith was leading. And so, for years, that basically just continued to grow. So, I led the software and applications, the software and algorithms, that’s inclusive of error correction, quantum algorithms, the software stack inclusive of programming languages for quantum computers, the compilers.
Through that process, we did a lot of discovery of new algorithms, new quantum algorithms, new applications, we did a lot of resource estimation to understand what would these algorithms cost as they scale up, what’s required of the hardware. And then, that would feed back to the hardware team to say, “Here’s how many qubits we’re going to need.” Also, to what’s now the architecture team, to actually design the quantum machine to handle that kind of load, those kinds of applications. And then, we built the software stack with compilers, languages, as well as simulators, and studied and developed new error correction that enables these applications to be run reliably and successfully. So, I led all of that work. And through that, I should say, we stood up the Azure Quantum platform, which brought the first quantum computers to the Azure cloud.
And we brought forward a handful, including Quantinuum, IonQ, Rigetti, QCI, and Pasqal. And now, we’ve extended that. We’ve now recently, actually just last week, announced that we have the Azure Quantum Compute platform that brings the first reliable quantum computing platform forward. This brings logical qubits, right? So, logical qubits are better than physical qubits. It’s what we need to reach scale, to reach reliable solutions from our quantum computers. And so, we announced just last week the Azure Quantum Compute platform that uses our qubit virtualization system. This is how we do error correction, how we light up the logical qubits from the physical hardware. It uses the qubit virtualization system, and we tailor it to industry-leading partners’ hardware planes. This includes Quantinuum, which is an ion trap system, and then also Atom Computing, which is a neutral atom-based system.
Probably a lot of readers of this interview won’t know about Azure. That goes back quite a ways, but not specifically to quantum. Can you explain that a little bit?
Yeah, Azure is Microsoft’s cloud. You can think of it like the world’s computer. It’s a combination of CPUs and GPUs, services, platforms, capabilities, but all accessible through the cloud and then, integrated across. And this really brings forward for the work we’re doing in quantum computing. It’s really critical, absolutely critical that quantum computers are integrated in the cloud. And so, Azure is a core capability that’s needed for quantum computing. Quantum computers don’t operate along. They don’t operate without classical computing, and they need a lot of classical computing. Both, actually, for the error correction itself, to keep the system noise-free, but also, they need it for the applications, right?
The applications you run require an extensive amount of classical pre-processing and post-processing, so you really want to be integrated with high-performance computing. So, here, Azure provides that cloud-based high-performance computing. It enables us to, for example, explore vast reaction networks for a chemistry problem, helps us identify which of those configurations we should ask the quantum computer to run to calculate. The quantum computer then provides really high accuracy for that calculation as it scales. So, today, classical computers can still do this calculation, but as you scale to thousands of logical qubits, this will be critical, right? Quantum computers will provide you high-accuracy energy estimates, for example.
And then, you’ll be able to take that information and feed it to a classical AI model that can then train with that information, and now you get a fast, more accurate model that you can ask, say, quantum properties of, estimate, predict with. One way to think about a quantum computer is, you have it integrated alongside HPC and AI, but you can also think of it as a data-generation capability. To get that same data with classical computers alone, it would require billions of years or lifetime of the universe timescales, right? We cannot get this kind of accurate information from classical computers alone for certain problems, especially in chemistry and materials science.
At what point timeline-wise did you feel confident, or did Microsoft feel confident, that it was time to bring quantum computing into Azure?
That was, I want to say, seven, eight years ago. We brought the first quantum computers to Azure in the Azure Quantum platform, and that started, again, seven to eight years ago. As Azure came up and expanded, it became very apparent that cloud is how you’re going to compute going forward. It provides the most secure environment, a unified environment, provides integration across services, it’s where your data can be housed securely, it’s where you can leverage now, for example, these AI models.
So, it is where everything is sitting. That’s not to say it can’t operate also with on-prem capabilities. But the cloud provides computational resources and data that you cannot otherwise have. And quantum computers need that information. It needs that pre-processing, it needs to integrate with these capabilities. If it sits alone, you’re not going to leverage its capabilities. You’re not going to be bringing together the best of compute. So, the cloud needs the best of compute. The cloud needs the best capabilities. And the best classical capabilities are in the cloud. So, you have to bring the two together. They have to be integrated seamlessly. Cloud compute didn’t exist 20 years ago the way it does today. [laughs]
So, it was really kind of in that, I don’t know, 2015, 2016, somewhere in there, that it became apparent, “Okay, the cloud is large. There’s a lot of computational capability there, and we need to make sure this integrates with it.” And so, we started to think about hybrid applications around that time period, but not only that, we were looking at chemistry problems, we were looking at materials science problems. You need classical compute to help pre-process that information as well. And so, we started to understand, “Okay, the structure of applications for quantum computers has a lot of cloud computing as a part of it, so we need to integrate it.” And so, that’s when we really realized it’s critical that we are a part of Azure.
You’ve developed the Q# [“Q-sharp”] language, so that’s all part of this as well, and the Quantum Development Kit. I think there were a lot of announcements in the 2019 period. It seems like a big turning point for you.
That’s right. So, around that time is when we really shared that we had the first quantum computers coming into Azure. We also had a Quantum Development Kit with which one could program those quantum computers, a number of samples, a number of learning exercises called the Quantum Katas, and all of these things remain available today. And now, we’ve augmented that with a Copilot for Azure Quantum, where that Copilot can help in your learning as well. It can help you with programming a quantum computer, it can help you write code, but also, it brings together multiple disciplines. So, the Copilot in Azure Quantum now is grounded in chemistry and materials science as well as quantum computing. This is where we really want to see cross-pollination of ideas. So, the Copilot, you can ask it about chemistry, and materials science, and problems in that space as well as problems in quantum computing.
You mentioned also some of the partnerships that you’ve had with IonQ, Quantinuum, Rigetti. Can you explain a little bit in more depth what they bring to all of this?
Right, so in Azure Quantum, we have a number of industry-leading partners that are a part of the Azure Quantum platform. This means that you can access those quantum computers through Azure Quantum, you can program those quantum computers. So, quantum computers are real, they exist today, and they’re available through Azure. So, you can use them, program them, run a number of samples that we have available, sample algorithms and applications, on them, and really explore and understand what you can do with quantum computers. Now, with that said, today’s quantum computers, up until what we’ve recently shared, have been what we call noisy, intermediate-scale quantum computers. So, you’re accessing these quantum computers directly to the physical qubits.
When you’re programming a NISQ machine, which these we’ve just talked about are, those are physical qubits. And quantum computers have noise. So, what we want to do is, we want to bring these physical qubits to the level of logical qubits. We want to orchestrate across the physical qubits to make them even better. So, we use them kind of as a pool of physical qubits to create a logical qubit. These logical qubits have better computational capabilities, so they have better error rates. You want to start to build algorithms on logical qubits. It’s why we invented Q#. Q# is a language designed to operate more at scale, to enable you to write algorithms that scale up, and it really talks to logical qubits. It can also talk to physical qubits, but the vision there is to start to program for scale. Today, we are now in the era of reliable quantum computing, so we have entered level two of quantum computing. Level two is the resilient level. Here, you’re operating on logical qubits.
These logical qubits need to be better than their physical qubit counterparts. And it’s at around 100 logical qubits where you can see what we call scientific quantum advantage. Scientific quantum advantage is when you’re able to have a solution from a quantum computer that you can’t otherwise get classically for a useful, interesting scientific problem. So, this is when a quantum computer outperforms a classical one. And this is a really important milestone, this is the milestone that we are aiming for with logical qubits. We’ve introduced the Azure Quantum Compute platform. We have two partners in that platform, Quantinuum and Atom Computing. With Quantinuum, we now have the most logical qubits on record, so on their H2 ion trap system, we have demonstrated together 12 logical qubits. For scientific quantum advantage, we want to drive that to 100 logical qubits. We’ve also announced a partnership with Atom Computing where we are working together to build a reliable quantum machine. We have a commercial offering around this, a discovery suite that includes this reliable quantum machine, HPC, and AI.
And together, we are working on building a machine that will achieve scientific quantum advantage. But also, Atom Computing as a part of the Azure Quantum Compute platform. And this Azure Quantum Compute platform, again, is, for industry-leading partners, we want to enable the best logical qubits we can from that hardware and our qubit virtualization system that’s in that Azure Quantum Compute platform. Now, we also want to advance, then, beyond level two to level three. It’s at level three where you have industrial quantum advantage, commercial problem advantage for problems in, say, chemistry.
Catalysis problems, problems that will help us better produce a fertilizer, problems where solutions will help us extract carbon from the air. These more complex, more advanced chemistry problems require a quantum computer that has over 1,000 logical qubits. We call that level-three scale, and it will require continuing to advance to level three to see this industrial quantum advantage. We also have a number of partners we’re working with as well as our own topological qubit, also, to get to level two and then beyond to level three. So, really, our intent is to have the most reliable quantum computing platform, and that’s Azure Quantum.
I had some interesting conversations last week about the topic of benchmarking. You’ve talked a little bit about different kinds of quantum advantage. There’s John Preskill’s term, quantum supremacy. I’m just kind of curious what Microsoft’s view is on those concepts, if you can describe it in an abstract or philosophical kind of way, in setting your agenda.
I use the term quantum advantage. This is when it’s advantageous to have a quantum computer in the mix. What does that mean? When is it really advantageous? Well, that quantum computer should not take 100 years to give you the solution. Not so advantageous, right? I won’t be here for that solution. So, we want to make sure that you get the solution in a reasonable amount of time. And it turns out, reasonable tends to be 30 days or less. It turns out classically at companies, when I worked at machine learning in Microsoft, we would deploy a calculation for 30 days on the compute that we had, so that’s very reasonable. So, when we think about quantum advantage, we really want to be looking at solutions. We want to ensure that the solution is achievable within about 30 days, and we want to ensure that it’s an interesting and useful problem.
It needs to be a meaningful problem. It needs to have value. It needs to be valuable either for me as a scientist, me as a business or enterprise, so it needs to have value. When I say scientific quantum advantage, I’m looking for when a quantum computer can give me a solution I can’t get classically, but I get that solution in a reasonable amount of time, and it’s for a problem that is of interest to scientists, to advancing science. And when I say commercial or industrial quantum advantage, it has to be useful for an industry. When will it be very useful for a given industry? And so, that’s more of an enterprise-grade solution, typically a larger solution space, more logical qubits required there, upwards of 1,000. So, there, we’re really looking at, for example, a catalysis type of solution.
I wonder if you could speak briefly about the relative roles of Microsoft, or industry generally, and the federal government. Of course, we’ve had the National Quantum Initiative in 2018. I know that you’re part of the Northwest Quantum Nexus, Pacific Northwest National Lab was part of that. And I know that you’ve recently been working with DARPA on utility-scale quantum computing, so it seems like there’s kind an ongoing set of partnerships there.
Yeah, I think it’s so important that we work together to advance this technology. It’s definitely the case that it’s not going to be just one institution, one company that’s going to have all the answers or all the solutions. It takes a world to advance this technology. and I like to say, it takes our collective genius, right? It takes people from different backgrounds, different domains, diverse people, to advance. And that includes when we look at the collaboration between government, academic institutions and universities, and industry, right? It really takes all three. We need to help educate and inspire a generation of quantum mechanics, of quantum computer scientists, of folks that are going to use and come up with new applications of this hybrid world, this hybrid HPC and quantum coming together.
And so, governments around the world, our government here in the US, play a really critical role. I sit on the National Quantum Initiative Advisory Committee, so it’s really important that we come together and collaborate. And there are a number of ways to do that, everywhere from workforce development and education opportunities, but also, as you mentioned, opportunities to engage together. Microsoft is a part of the US2QC, this utility-scale quantum computing program that DARPA is running to look at, “How do we architect and design to build for scale to reach thousands of logical qubits and beyond?” It’s critical we get there. And so, we are heavily invested in our topological program as a path to utility scale, for example. And so, we definitely work with the government, everything from sitting on the advisory committee to working with programs, in particular, to accelerate this technology forward so that we can empower people around the world with it.
Quantum information science is still kind of a heavily male-dominated field. You’ve been in it a long time. I wonder if you could talk about your experiences as a woman in the field, how you’ve seen that change or not change over time.
Absolutely. It’s so important that we have a diverse workforce. And in particular, I can speak as a woman, as a female in the field, there’s definitely more diversity needed across the board, and in other directions as well. Gender is just one of many. But of course, as a woman in the field, I would love to see more young women, women in general, but especially the younger generation, entering this field. I think that there’s so much opportunity to learn from folks coming from different angles, different backgrounds. I had the remarkable opportunity to have a number of female role models, and I think if you don’t see someone like you in a role, it’s hard to imagine you can ever have that role.
And so, it’s why I think it’s so important to have female role models and mentors, people that can inspire you, so that you think, “Okay, I could do that. She’s doing that, I can do that.” I had the opportunity… I mentioned already, Barbara Terhal at IBM, amazing, amazing woman. I am so incredibly grateful for her. To be able to see that she was in that role gave me an opportunity to think, “I might be able to be in a role like that someday.” And it wasn’t just her, I had Ingrid Daubechies at Princeton. I just ran into her, we’re both on the Flatiron Science Advisory Board. Again, she was in mathematics at Princeton 25 years ago. There were very few women. She was there. She encouraged me to stay in mathematics. There were two women, me and one other woman that graduated that year in mathematics. And it’s thanks to Ingrid that I really stayed.
Again, I think it’s really important to have female role models. I am committed to staying in this field for that reason. I want to make sure that there are role models for more women to enter. But it starts at a very young age. We need to encourage mathematical and computational thinking for all people, for all young people. We should not imply that math is hard, right? We should encourage looking at math in new ways, computational thinking in new ways. It’s important to understand. Like, cooking is chemistry, right? Science is everywhere.
And that can really turn on a lightbulb, can shine a light on an opportunity for a young person. So, I think there are new ways we can think about educating and inspiring, especially young girls, around the world in this space. Physics is fun. It’s really fun. And math is fun. And how it comes together. You might not realize, but you’re using it all the time. So, I really do hope we can change the outlook. I used to have a sticker on my door that said 50/50 by 2020. That didn’t work out, but I’m still committed. We’ll get there.
Well, thank you very much, Krysta. That does bring us up to about an hour, so I really appreciate your time today, and hopefully we can maybe even speak again at some point in the future. But this is great. Thank you.
Thank you, Will.
[End]