Summary
Simon Phillips, Chief Technology Officer at Oxford Quantum Circuits is interviewed by Yuval Boger. Simon and Yuval talk about OQC’s unique approach to building quantum computers using superconductors, why OQC’s quantum computer is online just some of the time, Simon’s journey from gaming to quantum computing, and much more.
Key takeaways
- OQC built its qubits around a coaxmon design, a coaxial-cable-based architecture that differs from the standard transmon approach used by most superconducting quantum computing companies
- OQC's quantum computer is not available around the clock because the cryostat and control systems need scheduled downtime between jobs, so uptime is managed rather than continuous
- Simon Phillips moved into quantum computing from a background in gaming, bringing an engineering perspective focused on systems and user-facing reliability rather than pure physics research
- Running a superconducting quantum computer involves real operational tradeoffs between keeping the dilution refrigerator cold and available versus cycling it for maintenance and calibration
Transcript
Yuval Boger: Hello, Simon, and thanks for joining me today.
Simon Phillips: Hi there. It's a pleasure to be here.
Yuval: So who are you and what do you do?
Simon: So I'm Simon. I'm the CTO, Chief Technology Officer, at Oxford Quantum Circuits in the UK. And, yeah, I've probably got one of the more exciting jobs in the world.
My role is to be entirely responsible for the technical output of Oxford Quantum Circuits. We're a quantum hardware company. We build quantum computers based on superconducting circuits, and my job really is to set the goals and vision of the technology, and then work with all of our internal technology teams to deliver against our roadmap.
That overlaps into the world of physics, the world of product development, the world of finance, and all of the things that we need to do as a company to make sure we can deliver the world's best quantum computers.
Yuval: I agree with you that CTO is probably one of the best jobs out there.
Simon: It's cool, right?
Yuval: You mentioned superconducting qubits. Now, I think other vendors are using superconducting for their quantum computers. Could you explain what makes Oxford Quantum different?
Simon: Yeah, absolutely. So we're based on superconducting circuits, and it's quite a popular modality, if you like, because of the engineering capabilities and some of the routes to long-term scaling.
But at Oxford Quantum Circuits, we've got our core technology. It's called the Coaxmon, and this is what we spun out of the University of Oxford with in 2017. It's a 3D architecture for superconducting circuits. It's completely 3D inasmuch as our control wires are completely out of plane.
So we have a really beautifully clean substrate with our qubits and resonators on it, and all of the control wiring and all of the bits that the qubits hate, that bring interference and crosstalk, we've taken out of plane, and they come in completely 3D. That gives us a number of really good advantages, or fewer challenges to scaling than some of the other superconducting circuits.
Yuval: Are these challenges just related to noise, or do you believe you'll be able to deliver, for instance, better connectivity or a higher number of qubits sooner?
Simon: Yeah, so—
Yuval: The marketing spin.
Simon: Yeah, so without getting too marketing and spinny, what it really allows is, if you imagine taking away all that control wiring from the substrate of your chip, there are no through-silicon vias or air bridges or bonding.
The other really cool thing with the Coaxmon is we don't ever galvanically bond to the chip, so it sort of floats in space. What that means is we're completely free to run any kind of connectivity pattern between the qubits without having to navigate any wires or re-architect anything. So it gives us a really clean, neat solution for trying some really exciting things.
But ultimately what it means is there are fewer things to interfere with the qubits. And because we come straight down in 3D from top to bottom, in a capacitive coupling style, we have really low crosstalk between our qubits. They're perfectly mode-matched up and down.
So as we scale the number of qubits, we're really not compounding crosstalk or other kinds of errors. We're still noisy, and we're all trying to work out the best and neatest way of doing error correction as we move towards being fault tolerant.
But for sure we start with a lot fewer challenges than other people, which gives us a lot more freedom for design.
Yuval: And given that it's superconducting, you still need cooling, right? You still need cryogenic cooling, or very, very low temperatures.
Simon: Yeah, that's right. We run at millikelvin levels, as all superconductors do.
But again, the really neat thing is, because of the 3D nature, as we add qubits to our processor, we get linearly bigger, right? We're adding rows and columns of qubits, as opposed to getting exponentially physically bigger with 2D wiring or bump-bonding stacks and things like that.
So even though we have cryogenic cooling, the physical space and the density of our processors means we can fit many more qubits in an off-the-shelf dilution refrigerator. We're quite fortunate in that, with the largest dilution refrigerators you can buy off the shelf today, we can see really far into the future with that infrastructure.
So again, it's one less thing to worry about, right?
Yuval: To the extent you can share, what is the stage of the company in terms of funding, number of people, working product? Where are you in the quantum journey?
Simon: Yeah. So curiously, we still call ourselves a startup. I'm not quite sure when that changes, to be quite honest.
Back in the day, in a previous life, I used to make video games, and a startup really meant you're sat in a bedroom with your mate, coding video games. But we've just done a first close on our Series A. It was a £38 million round, which we're really proud of, and the largest quantum Series A in the UK.
That was pretty cool. We've also commercialized our current-gen quantum computer, Lucy, an eight-qubit quantum computer, and we launched that over Amazon Braket back in February, I think. That was a really big milestone for the company, to do the thing we want to do, which is accessibility, right?
The more people that can use quantum computers, the better people can learn what they're doing. The barriers to entry come down for business value problems, and the whole ecosystem starts building. So we've launched our first quantum computer. We've just closed our Series A. As part of that, and this is public information as well, we've got a new co-lead investor that's going to help us internationalize the company.
So we're starting to look at what we can do in other territories. We've got a Japanese co-lead investor, so that's really exciting for us, to leverage talent in Japan and see what we can do with quantum computing there as well. And now we're able to really take advantage of the Coaxmon core technology in terms of scaling.
So we're going to see quite a quick increase in our qubit count, but more important than just qubits are our quantum volume and our error rates as well. We want to focus on high-quality processors rather than just chasing qubit numbers. But with this fundraise, we should now see real quick acceleration through the gears.
Yuval: Let's dive a little bit deeper into the Braket integration, because I'm curious about several aspects there. One thing I think I saw in the press release is that now Amazon Braket has a computer that's in Europe. If I'm an end user, why do I care where the quantum computer is?
Simon: Right. Yeah, absolutely. So AWS in Europe is our quantum computer. It's Lucy, OQC's quantum computer. And it's important for a number of reasons, actually. First and foremost, the really big drive at the moment is around data sovereignty, or data onshoring. If you're starting to look at building proprietary algorithms or applications, you want to make sure that you've got complete data traceability from where you start to where you finish.
You really want to know where that computer's located. That's been a really important requirement for a lot of people we've spoken to. So there's that aspect to things. Next, we live in a world today where we have uptime for quantum computers. We run in windows—not the operating system, but windows as in gaps of time—and we need to schedule that time to be friendly with academic research groups or businesses or things like that.
Otherwise we end up queuing, and we're quite passionate about how we can move that forward. So there are a number of reasons why it's important, and going forward, latency is going to be a really big issue as well. Today it's not so much of a problem because we're only dealing with handfuls of qubits, but fast-forward a few years, when we've got hybrid applications running on CPUs, GPUs, and QPUs, and you're really going to care about latency between where you're based and where the other technology is based.
So that's primarily why it's important to be in Europe.
Yuval: Regarding the uptime, why is it not up 24 hours a day? I mean, I know no quantum computer is up 24 hours a day, but do you have to tweak the qubits in between operations? Do you have to clean things up? Are you making changes to the machine as you go?
What's the reason?
Simon: So there are a number of things that thread into it. At the moment we live in a world where we're constantly calibrating the computer. We're really learning about what drift means for us and how we can constantly keep the results better. And bear in mind, we build the whole back-end stack, right, from fabricating qubits.
We build our own room-temperature control electronics. We've got our own software stack with compilers and optimizers. So we're constantly able to squeeze more performance and power out of the same hardware just by keeping the thing well calibrated, or up to the latest firmware designs and things like that.
So we are constantly tinkering with it. And you're right to say it's on 24/7, right? The cryogenics are always on; the qubits are always there. The uptime at the moment is really based around the way the partnerships are forming with people like AWS, in that they want, and rightly so, really good SLAs for customers.
So if you're going to run something on our computer over Amazon Braket, you need a certain level of service level agreement, where you know that if you've got a problem, it can be solved on time. If you hit a problem with a quantum computer, we have someone on it within 15 minutes to solve that problem.
But what that really means is we end up narrowing down to an intense window of operation. Our vision for quantum computing, though, is one of seamlessness in the future, right? Much like today, you can almost take for granted that if you want 1,000 GPUs on your network, you can just provision them with your local provider, or if you want more storage or more compute power, you can just add it from a menu and it's on your network within seconds. That's where we see quantum computing going.
So we're talking about always-on, always-calibrated, always-available resources. We've got a long way to go, of course, before this is realized, but we really want to move away from this kind of uptime and windows to just being able to spin up a QPU like you would anything else. That's the goal.
Yuval: If I'm an Amazon Braket customer, I could run it on Lucy, I could probably run it on Rigetti or IonQ or a number of other vendors. And let's say that I find your technology very cool, but as a customer I just care about the results. Why would I choose Lucy over other computers that I could use on Amazon Braket?
Simon: Yeah, that's a great question, and we fully embrace that ability on Braket. One nice thing is you can write your code in Braket and just swap out the ARN at the top, right? So you can jump from Rigetti to OQC, and I think that's really cool. For us, what's important is that we've got high-quality processors.
So long coherence times, longer coherence times, with good one-qubit and two-qubit fidelities. We like to think that the results you get from Lucy are going to be really high quality. That's one thing that's really important. And one thing we're super keen to do is really support the community in understanding those results as well, right?
So we have a lot of ways of communicating via Amazon Braket, whereby if you want to understand your results a bit better, you can get in touch. We can make postings and really help you understand how to write code for quantum computers. And I think that's a really important point, because we have to acknowledge that today the computers aren't useful for business value problems.
Today it's about learning how to use them and learning how to think about developing quantum algorithms or quantum code, and it's great to be able to do it at a level where you can write some assembly language for eight qubits and really read it, right? You can understand the code, you can understand the results, and you can start mapping that to your learning process.
As the processors get more complicated, you're actually going to lose that ability to fully understand what's going on. So we encourage people to look at the results, reach out to us about what they mean, and we really help you understand what it is to write code for our processors.
Yuval: Some companies believe in vertical integration, that the software should be really tightly coupled to the hardware to take maximum advantage of it. Whereas, as you describe, on AWS I could change a couple of lines in my quantum code and run it on yours versus someone else's. How important to you is that coupling between software and hardware—the optimization of the software to take maximum advantage of the hardware and overcome its limitations?
Simon: Yeah. So I think that goes up and down the stack a little bit in different ways. We certainly believe in it when it comes to the control of the qubits themselves. This is why we started a program three or four years ago to build our own control electronics for our computers. We have a team of people dedicated to PCB design, FPGA design, and actually tuning the control hardware to the topology of the processor itself.
So as we move into error detection, error correction, and things like that, we've got really high throughput of information. When you run a task, it's going to run in lightning time. We can then do all the things we need to do to build our error correction without having to buy off-the-shelf hardware.
It's all custom designed. So we believe that needs to be really tightly coupled. Then, and I guess most people do this, we have our own compiler, which will take advantage of the nature of the processor itself. So you can write more generic code, and our compiler and optimizers will map that to the processor in the most efficient way.
That's great for getting good results. It doesn't always help learning, though. Sometimes you just want to run it verbatim, right? You want to say, "Run on qubit 0. I just need to see how these qubits are going to operate." So we offer that kind of flexibility. But I think we really want to get to a world where, above that, in the application space and the algorithm space, you can use whatever tools suit the person or the problem better, and then let the rest of the stack, from our software through to our hardware, do the tight coupling for performance.
Yuval: Can you give us a little peek into your technical roadmap?
What do you expect Oxford Quantum to offer in the next year or couple of years? Is it just—
Simon: Yeah, it's like—
Yuval: More qubits, or is it something else?
Simon: There's a multitude of things. There's always going to be more qubits. But like I said, what we really want to do is not necessarily focus on the number of qubits.
We've actually built our roadmap around target error rates. I can't go into the specifics of what they are, but we really feel that the number of qubits matters to a point, and then the more important factor becomes the fidelities and the error rates that we're able to achieve, and we map that against algorithms and business use cases from there.
But I think we can think about our roadmap in three different strands. We've got the main technical strand, the processor power. That's one thing. Then we've got what we call customer on-ramp, and this falls firmly into the technical space as well: how can we get end users to write code for the quantum computer that starts solving their problems?
It might be specific features that they need, it might be specific languages that they want us to ingest, or it might be a combination of those things. But really we need to help enable large businesses to build out the teams of people who will develop algorithms that are proprietary to them.
So we've got a big strand there. And the third piece for us is infrastructure. We need to make sure that when the processors are capable enough and the customers know how to build their quantum algorithms, the computers are where they need them to be in terms of data sovereignty, connectivity, hybridization, and all those things.
So we're going to see some quite exciting things in all three of those strands from OQC: processor capability, customer interactions and customer on-ramping, and infrastructure as well.
Yuval: People get into quantum computing from different areas, and I don't remember hearing about anyone else who came to quantum computing from gaming.
Simon: No.
Yuval: How did that teleportation process happen?
Simon: Yeah, it's a really interesting one, actually. I started in the video games industry when I was quite young, in the '90s, let's just say. And one other thing I did enjoy there, as I said earlier, was being an assembly programmer. Just being able to read assembly language and know what it's doing at a hardware level is really fascinating.
I'm quite enjoying looking at QASM code now. And really, that journey for me was one of enjoying technology and enjoying business and entrepreneurship. We did some really good things in the video games industry in terms of growing businesses, and real cutting-edge stuff.
Well, cutting-edge maybe not in the same way that deep technology is, but we experienced when 3D graphics first came out, when free-to-play business models first came out, when the internet first came out, and all these big firsts where it was like, wow, we have no idea what's coming next, we just have to architect the future, right?
And I see a lot of similarities in quantum today, whereby no one really knows how this business is going to evolve or how the business model's going to evolve, and it's so exciting to start shaping that. But yeah, I've always loved technology. I have a bit of a cheesy one-liner that a lot of our current investors will have heard me say in a pitch: I thought I could change the world with video games, but really we needed quantum computers to do that.
So fortunately for me, Ilana really bought into my passion for technology and my passion for enabling and inspiring teams of experts. I'm really fortunate in all of the leads that we have. We have a number of technology teams within OQC, and they've been set up in a very specific way to enable all the leads to have full ownership of their space.
These are world experts in quantum, or in decoherence, or in fabrication and things like that. So my job really is just to enable them to think big and to have the resources they need to do things, and that's everything I learned in the video games industry.
But for sure, I'm in a really fortunate position at the moment.
Yuval: As we get close to the end of our conversation, I'm curious: what can you tell me about the UK quantum ecosystem, and maybe, if you know, how it compares to, say, France or Germany or other European countries?
Simon: Yes. Actually, North America aside, which has got a pretty good ecosystem, I've seen it change a lot even in the last two or three years.
Three or four years ago you'd say you're in quantum and people would look at you like you'd just walked off a spaceship. Whereas now, I think we've grown from… well, I think I was employee number four or five or something like that, and we're now at 65 people, I think, today.
And it's only now that we're starting to see people who have industry experience in quantum, right? A few years ago, either you had a PhD, or we were bringing you in from a different field entirely and upskilling you. And really, it's only now that this ecosystem has got to a point where there is experience there, and that's fantastic.
That's the only way this whole technology's going to thrive, and we're starting to see that in other areas. I think the UK feels relatively advanced. Europe's great from a financial point of view. They seem to be able to get a lot more European money into businesses there, which we're not jealous of at all.
But it's definitely growing up. It's definitely growing fast. And part of our goal, again, is to help stimulate that ecosystem. I think the bigger it is, the easier it is for all of us to find people, find connections, and be ready for quantum.
Yuval: And Simon, how can people get in touch with you to learn more about your work?
Simon: I'm more than happy for people to drop me an email. I love chatting about technology. I love chatting about science. I love chatting about quantum interaction. So people can reach out to me at simon@oxfordquantumcircuits.com. It's a bit of a mouthful, that URL, but yeah, just simon@oxfordquantumcircuits.com, and I'm more than happy to have chats.
Yuval: Well, it's been wonderful chatting with you today. Thanks so much for joining me.
Simon: Brilliant. Thanks for inviting me. It's always good fun.
