Transcript | Warm, Interconnected Qubits— with Universal Quantum

Scaling quantum computers would be much easier with high-fidelity interconnect and higher-temperature qubits. Find out how one company is building exactly such a system in this episode of The Post-Quantum World. I’m your host, Konstantinos Karagiannis. I lead Quantum Computing Services at Protiviti, where we’re helping companies prepare for the benefits and threats of this exploding field. I hope you’ll join each episode as we explore the technology and business impacts of this post-quantum era.

Our guest today is the CEO of Universal Quantum, Sebastian Weidt. Welcome to the show.

Hi, Konstantinos. Thanks for having me.

For everyone listening, we can start out by exploring how you got involved in the quantum field.

I don’t have one of those nice, inspirational stories to tell where I dreamed about quantum when I was five years old. That’s not quite me. I always wanted to work on something that has impact, and back in the days when I was younger, I thought that would be through business. That was quite clear to me at the beginning. But I also recognised that I probably have to show that I can do some logical thinking, and what better way to do that than study physics at undergrad?

I actually did physics with management, and that’s where I started to get some exposure to quantum. But I stuck to my guns and went to management consulting for a little bit, but then decided to do a Ph.D., and I ended up doing a Ph.D. in quantum, focusing on quantum computing, and helped develop some nice things and started to realise that quantum computing isn’t just confined to academia, but that it’s something that would take some commercial trajectory and that the impact, besides, is unbeatable.

That got me excited. I stayed on, did a postdoc and followed the academic track. I initially became a professor in quantum computing. But at some point, we realised in academia that you can’t build the sort of quantum computers we have in mind in that will help us unlock these nice applications of quantum computing in academia. It felt like it ought to be a venture that’s more commercially focused to attract the right sort of people and attract the right sort of money you need. That’s how I ended up in this field, in a way.

In academia, a lot of times, you end up getting this paper that gets published with an approach and then, “No, now we’ve got to build a company around it, because we’re not going to be able to turn this into a machine.” It’s funny, to point out something you said, that you don’t have one of these inspirational stories, but you do: You came to it a little later, and you were able to still dive in and become part of it.

For a lot of our listeners, they’re interested in the field and want to make the switch. It is always fun to hear about someone who maybe a little later decides they want to switch over. Great story. Tell us about — back to that point of, in the private sector, you have to have a company. Tell us a little bit about Universal Quantum and its aims and goals.

Universal Quantum is a spinout from the University of Sussex. We are in the U.K., and it focuses on building quantum computers — but not just any quantum computer. What we as a company are focused on is providing humanity with a tool to do something amazing. When you apply that thinking to quantum computing — and I’m sure we’re going to talk about that later on as well — you start realising the sort of quantum computer you need to build needs to be a pretty scary-big machine: lots and lots of high-quality qubits.

That’s what we’re passionate about at Universal Quantum. We’re a company that focuses on scaling up quantum computing, particularly focused on trapped ions. Trapped ions, we use as our qubits. Over the many years in academia and at Universal Quantum, we’ve managed to remove some of those major obstacles associated with scaling up this particular technology.

The other thing worth pointing out is that with that bold vision — some people would say building very large-scale quantum computers, you naturally have quite an engineering focus. If you look at Universal Quantum, you won’t just find lots and lots of physicists running around. I’ve been very open about my belief that you don’t just want to have physicists build a quantum computer. It ought to be engineering-led. It needs to be some of the world’s best engineers together with those amazing physicists building this. This is why we’re very engineering-focused. If you look at our team, it’s heavily leaning toward the engineering side, which, as a physicist, I had to get used to, but I’m a strong believer in that approach.

One thing we can agree about with physicists is, to them, every cow is a sphere. Maybe we shouldn’t be thinking along those lines for manufacturing quantum computers — you have to get down to the engineering practicalities. Let’s start with the biggest goal of the company, and it’s a modest goal — a tiny one: to build a small, little, teensy-weensy, million-qubit quantum computer.

That’s a pretty big goal. Why not start with a thousand or something on par with the systems that are coming out this year?

One thing, just to be very clear: We’re not saying that the first machine we’re going to build is a million-qubit quantum computer, and you can pick this up next year. It’s pointing to what is ultimately required. It’s important to have that conversation around what quantum computers ultimately need to look like to unlock these amazing applications we associate with quantum computing. I don’t think 100 or 1,000 qubits will be enough. It ought to be more, long-term —millions of qubits.

Why do we say this right now? This is, again, where maybe Universal Quantum takes a bit of a different tack, where we’re pretty worthless with ourselves in a way, when we develop our machines, where we always keep that eye on the millions of qubits, where everything we develop, even for the first machines we’re developing at the moment, it needs to withstand that question of, “Does that scale to millions of qubits?” Thereby, the first machines, which will have far fewer qubits, will look like a mini version of a million-qubit quantum computer. That is, for us, the right approach.

To put this a different way, let’s say we solely focused on building a 1,000-qubit quantum computer, and our engineers only have that goal in mind. We get to the 1,000-qubit level, we’re going to be incredibly excited and happy. But then, obviously, immediately, the question would be, “I’d like to have 2,000, please. I’d like to have 10,000 as a next step.” Well, what the engineers and everyone involved in building these machines will then ultimately say is, “But now, I need to walk back and redevelop a lot of the things I’ve just developed, because I never thought about 10,000 qubits.” So you start again.

And for us, it’s about the point where we have a 1,000-qubit machine, what we want to be in a position of is when we want to get to 2,000 or 10,0000 or 100,000 qubits, it’s just more of the same. That, for us, is important. It’s a bit more painful at the beginning because we’re doing a lot of heavy-lifting engineering work at the moment to get to a state where the solutions that we employ do scale up. But we do feel in the long term, that is the right approach to take.

I want to get into one of the biggest pillars of your approach. You broke it up into six in general — that’s available on the site. Scalability, I always feel, is greatly dependent on something like the ability to have modular systems that can be stacked and daisy-chained or whatever, built up — a concept we talk about here often. You are working on something called UQ Connect, which is a type of interconnect, which is critical. Without it, I don’t know how we can ever scale. This one landed a publication in Nature. If we could start talking about why it has a specific name like UQ Connect, as opposed to just another interconnect, maybe you can explain what that is and why it has a fancy moniker.

It’s great you’re picking up on this particular pillar because it’s absolutely crucial, like you say, to how we’re thinking about scaling up. You’ve mentioned already that if you do think about scalability, modularity is incredibly important. The idea of how you connect individual quantum computing modules together becomes a pretty important question. Now, I often fear that not many people have a good answer to that, and that there’s still a lot of work to be done there in many areas. Now, for us, because of the way how our architecture works, we asked the question about if there’s maybe a unique way of doing our particular way of connecting modules together that alleviates some of the pains that some other approaches are currently seeing.

What, in a way, we came up with high-level is a way to connect individual quantum computing modules using electric fields. Now, these electric fields, they’re naturally there all the time. It’s one of the ways trapped-ion quantum computing works. We’ve been able to engineer our modules in such a way that you simply bring them close enough together. This magical electric field link opens up so we can physically move trapped-ion qubits from one module to the next. There’s no additional engineering involved. There’s no intermediary qubit or photon or something like that involved. It’s as pure as it gets. The information never leaves our qubit. It just moves from one module over to the next.

You’ve mentioned this particular publication where we’ve demonstrated that. I remember when we proposed this idea many years ago, a lot of people thought we were somewhat crazy to try and do that — have these trapped-ion qubits hop, so to speak, from one module to the next. But that’s where we demonstrated — I’ll let you be the judge of the quality of the results there, but it’s working remarkably well to the point where we believe we don’t have that as a problem anymore. The question of how to connect modules, for us, that has been ticked off.

Do you feel that you have a sense of how many qubits should be in one of these modules yet before you start expanding?

If you would have asked me this question a few years ago, I would have said, “We’re going to make these modules as big as possible — push toward maybe wafer scale, and then squeeze as many qubits on and then make that step.” But what we’ve realised through the work we’ve done on these connections, and because these connections for us work so well, is that it no longer matters how big each module is, because for us, making a connection is pretty much error-free. It doesn’t take a penalty, and therefore, we can focus on letting the engineering be the driver of how big the modules should be — yield considerations. Generally, when you make these modules, there are things you have to think about purely from an engineering point of view. It’s no longer dictated from the quantum side.

Are there issues with the support circuitry for each module, though? If you were to have too many qubits, would it be too much interference in these modules? Each one is close together, and they’re each trying to control, and then, all of a sudden, it becomes that equivalent of having a lot of wires or a lot of circuitry. That’s what I was thinking along the lines of, is there a number where if you go too big, it becomes an issue?

Right now, across the board, the problem is, you add more qubits, you need more wires, and it just becomes a mess, and this is why we haven’t talked about that bit yet. The modular approach is asking for a good way to connect individual modules, but it also puts a requirement on what each module looks like. For this to work properly in modular architecture, each module needs to act like a stand-alone mini quantum computer. Most of the things you need to control the qubits in your module need to be integrated into that particular module, and that starts to get you around a lot of those wiring issues and so on.

The other bit, obviously, that you need is the way you control the quantum. We can get to that point later on as well. You’d better have a way to do that in a noninterfering way. Again, we have some novelties that allow us to achieve exactly that.

When you talk about an approach like Signal, that’s typically what I call hybrid post-quantum cryptography. We’re starting to see that rollout. Of course, the newest version of Chrome supports this kind of approach. AWS supports it within its own infrastructure there. It’s one of those times I like to say that being on the cloud is more secure than not as you approach this quantum era. But there are still either custom-written things or other paths that can’t be protected by these kinds of means. That would be where you start to look at custom solutions. We should roll into a few of the things your company does today while you’re waiting. Then I have a follow-up question based on the White House memorandum and some guidance. But we’ll start with basically what you provide companies today that want to start securing certain critical paths.Let’s move through the other pillars and get a sense of what these look like to visualise it. The first point to bring up is, as you said earlier, these are trapped-ion qubits. We’ve seen this before — a couple of other big companies making these. What kind of fidelities are you seeing, and what makes your approach different from, let’s say, the bow-tie approach we’ve seen before?

It’s probably fair to say that trapped-ion systems are the best-performing ones. If you look at the achievable gate fidelities, they’re world-leading. It’s one of the reasons we like this qubit. It’s well-isolated; it’s well-controllable. It’s beautiful, and the fidelities are amazing. But equally, when you look at how those quantum gates are implemented, it sometimes becomes challenging to see how that pristine control survives as you scale up.

Often, people like to use the individual laser beams aligned onto individual trapped-ion qubits, which works amazingly well at the moment. How that scales up remains to be seen. We had some concerns around that, in a way, where we developed a different gate technology which negates the use of laser beams.

Here, we’re getting into one of our other pillars, where we replaced the requirements of laser beams, where, to put this in the listener’s context, usually, you would need around two laser beams for every qubit you want to operate on in parallel, on order. It’s OK where we are right now, and it may be OK for 1,000 qubits, but as this scales up, one can imagine that gets pretty complicated. We’ve replaced the need for lasers with a handful of microwave fields.

The key thing here is not in the microwaves. I also wouldn’t want to have millions of microwave fields to control and scale. We took this another level further, whereby, no matter how many qubits we have in our quantum computer, we only ever need a handful of microwave fields that are being sent everywhere to all qubits. We developed a trick whereby we maintain the individual controllability of each qubit. That’s the nice thing when it comes to the control approach we have at Universal Quantum.

Is there some interference that happens with the fields to control particular qubits?

No, fortunately not, because people sometimes try to play around with interference. It gives you something, but it doesn’t work that well. We don’t use that at all.

We’ve basically developed two things. One, we have a qubit, which we can tune: The energy we need to put in to change a qubit, we can tune that qubit so different energies will be required to change the qubit state from 1 to 0, for example. That controllability allows us to tune a qubit into resonance or into a particular global channel we have, just like your radio operates. Let’s say you have three channels. One is, do nothing. Another channel is to a single-qubit gate. Another channel is to do a two-qubit gate. They all live in different frequencies— different frequencies are being radiated to the quantum computer. We now have a way to locally tune each qubit into whichever channel we would like it to tune into — to, for example, do a single-cubic gate or a two-cubic gate. That way, we get around the interference problem.

For us, microwaves are everywhere. Some microwave fields go straight through a qubit and don’t interact with it. It only interacts with the qubits we want it to interact with. It’s quite technical, but high-level. The key thing is, there’s no correlation between the number of qubits in your machine and the number of radiation fields we have to send in. That is the big win for us. Therefore, technically, if you came to look at our machines right now, the control boxes we have that generate these radiation fields that replace the lasers, it’s the same box we have right now that we’ll have for 1,000 qubits, 100,000 qubits or a million qubits. That is the essence of it.

Scalability is thought of throughout, everywhere.

You’ll hopefully see it everywhere.

The fidelity of these gates and qubits, do you have numbers for those, or is it too soon?

We’ve published some stuff in terms of what Universal Quantum put out there. Wait and see, and you get to play with it. But we’ve demonstrated that this particular gate we’ve developed works well. But what the first machines will show, I’ll refer you to when they come online.

We’ll talk about timeline a little later on. One other aspect of these qubits is, they’re something you don’t normally hear: They’re warmer than what you’d expect. Can you talk about that?

In general, whenever I talk to people about quantum computing, one of the first things they say is, “It’s really cold.” A lot of people become accustomed to having to cool down to many-Kelvin temperatures near absolute zero. Trapped-ion, in general, has this advantage that you don’t need to do that. There’ll be some systems that are sitting around room temperature. But in general, for us, for example, what we’d like to operate at is 70 Kelvin, which is a very, in the grand scheme of things, warm temperature, where a lot of cooling power is available even at scale.

This is the key part: If anyone wonders, “Why do you care about temperature?” When you’re very cold, you don’t have a lot of cooling power available. Generally, as you make your quantum computer bigger, you have to dump more energy in. That heats up the system, and it causes you some problems. Therefore, the warmer you are, the better these cooling systems work, and the more cooling power you have available to put more stuff in. At 70 Kelvin, we feel that that can take us all the way to the million-qubit scale.

That’s so many orders of magnitude warmer than other qubits. I don’t know how to come up with an analogy for that. It’s like standing on Earth and standing on the Sun — it’s like almost that much of a difference in terms of magnitude. We have all this working on these modules, and they’re fully integrated, so everything is inside per module. Can you talk about that design and what will make it so scalable? We got everything there. We have the microwaves. We have the whole gate system in place. How do they connect to something else like that — full integration? How does it get passed on to control circuitry?

The other part we need when we talk about controlling our qubits is quasi-static DC voltages that can change in time, and that allows us to physically move qubits around. Now, why do we like to do that? This nicely leads me on to talk about connectivity. What I mean by connectivity in quantum computing is the ability of a qubit to talk to another qubit in the system.

Now, when we’re talking about architectures where you’re physically printing qubits in place when they’re static in a solid, for example, you have a nearest-neighbor architecture where a particular qubit can only talk to the one that’s sitting right next to it. In our particular case, because we can move these qubits around in a pretty error-free way, we can get to a more fully connected system, which, high-level, means any algorithm you want to run runs so much better, so much more efficient. That connectivity speeds up the quantum computer quite significantly. Therefore, we need to think about, how do we generate, and how do we get these changing voltages, to the chip?

There, we do some integration work. Again, we develop specific ASICs — application-specific integrated circuits — we integrate into our chip architecture to generate these voltages right where they’re actually needed and no longer need these control wires you often see when you look at quantum computing systems: lots of wires always going to a very small chip, which is good right now but doesn’t scale. We need to get rid of the wires and integrate it into the chip. This is exactly what we’re doing with what we refer to as an integrated quantum processing unit, or IGPU. That’s another thing we’d like to integrate.

That makes perfect sense. When you start to look at all these potentials for things that could go wrong, all these controls that can go a little awry, we start to think about errors. I want to talk about what your thoughts are on error correction and suppression and mitigation, because they’re all different things. Which of those are you already experimenting with, and what are your plans? Maybe even some guesses on that dreaded ratio: How many qubits will it take to get a logical? Any thoughts around that area would be interesting.

First, obviously, aren’t these errors annoying? Wouldn’t it be nice if quantum computing didn’t have any? But at the same time, fortunately, error correction exists. For any listener who doesn’t know, the reason we’re talking about millions of qubits is because of the errors. Error-correction algorithms basically say, “Give me lots and lots of qubits, and I’ll remove the errors for you.” It’s one of the reasons we have to work so hard on the qubit-number side.

It probably won’t be a surprise to you, because we think about these larger-scale systems a lot — the point of error correction is a very big one for us right now, because if you want to start thinking about what a design for such a system looks like — and we published, for example, the world’s first blueprint to build a million-qubit quantum computer — you have to know what error correction demands, what it wants from you, how can it best operate so you can incorporate that into the design of your machine right from day one?

Some may say, “You’re thinking about it a bit too early,” but again, we’re weird, and we think it’s important to think about those things right now. Yes, we think about error correction a lot. We’ve got a number of collaborations and funded projects around that, and we’re trying to integrate what error correction requires into our modules right now. You also mentioned error suppression, error mitigation. They’re all part of the strategy toolbox we’re exploring to get a handle on these errors and are part of our ongoing development at the moment.

It might be too soon to figure this out, but do you think it would make sense to get the module to the point where it becomes a logical qubit or two itself? Is that the ideal module?

I imagine a lot of people have to think that way. Again, what I was saying earlier is, for us, because there’s no overhead to connect modules, we don’t have to work hard to try and get a logical qubit into one module. If you have a particular connect technology, which is lots of errors or very slow, you ought to think very hard about how you can maybe get enough qubits into one module to turn into a logical qubit, because there’s a lot going on with an illogical qubit that’s important. If you don’t have that worry on connecting modules, you’ve got some flexibility here, and we don’t have to necessarily squeeze that in. We can be a bit more free-thinking there.

That makes sense, because if it turns out that you need a thousand to make a logical qubit, that becomes trickier.

For us, obviously, one always wants to work on minimising this ratio that you talked about — the ratio of physical qubits to logical qubits. We’re doing some of our own work there as well to see if we can leverage some of the advantages our system has. We talked about connectivity and so on — if that lends itself to improve certain error-correction algorithms to reduce that ratio. But ultimately, we at least don’t have to worry about it from a hardware perspective to try and get this into the size of one module.

I’m going to have, in the show notes, a link to your company. I want people to look particularly at UQ Connect because just the visuals — just to get a sense of how it all works — it’s pretty cool to look at. Is it too soon to talk timeline, then? Do you have any sense of what the first time you’re going to slap a few of these modules together into a box might look like?

We’re building it all right now. It’s probably no surprise I won’t give you a date necessarily. But what is publicly available is, for example, not too long ago, we got two large contracts from the German government for €70 million to build them two quantum computers. The timelines there are publicly available — a bit slower than what we’re going to do elsewhere, but over the next couple of years, you’re going to see some pretty cool machines.

Did you give any thought to what it’ll be like to programme them? Are you going to build it on an existing platform? For example, will it be Qiskit going in or something like that? Or will you be developing an entirely new paradigm for that?

Programming is so important. There’s a lot to squeeze out of these machines purely from a programming side, and there’s a lot of work to do. There are some phenomenal people, companies, out there working on the software side that, some of them, we’re already collaborating with, and I wouldn’t want to try to replicate the amazing work they’ve done.

We remain very committed to having control over our software stack. We would like to maintain access to a customer directly. We are building a cloud infrastructure for that. But the way you can, in a way, imagine this software stack is quite modular, so that someone who’s developed a cool software piece can, in a way, integrate this into our modular software stack, and it can interact with the higher level and with the lower level. That’s, in a way, how we’re trying to marry the two concepts.

Like some other companies, you’re having your own cloud, but then maybe the potential for through some other abstraction layer, like connecting to Braket, to let people in. That makes sense to reach more potential users.

That’s exciting stuff. It’s the UQ Connect that is going to be the real game changer here, just from what I’ve seen so far, if I had to make a guess. I urge people to take a look.

Are there are any last thoughts you want to leave our listeners with before we sign off?

Look at UQ Connect, but look at the other things as well. When you think about scalability for us, you have to look at the whole architecture together because for us, it all comes together and works as one. But you’re right — UQ Connect is absolutely key. The way we control our qubits is important as well. Have a look. Engage.

Thank you for listening to the quantum podcast. It’s phenomenal how many people do listen to this topic. Thank you so much for listening to me as well. Please engage and reach out. If you have any questions, please get in touch with us. We’re very happy to speak to people out there.

Now, it’s time for Coherence, the quantum executive summary, where I take a moment to highlight some of the business impacts we discussed today in case things got too nerdy at times. Let’s recap.

Universal Quantum is a company with a plan to build a million-qubit quantum computer. They’re not counting on some future breakthrough to manifest, either. They’ve got a plan based on current technology.

For starters, there’s UQ Connect, a type of interconnect with better than 99.999993% connection fidelity. The technology allows qubits to move between independent modules freely if the modules are close together, all without additional engineering. I put a link to the Nature paper on this approach in the show notes. These moving qubits are trapped ions, and the control circuitry for them is integrated into each module. All this operates at about 70 Kelvin, which is several orders of magnitude warmer than most quantum computing modalities. Universal quantum hopes all these factors will contribute to a realistically scalable system, and the company is already contracted to build systems for the German government.

That does it for this episode. Thanks to Sebastian Weidt for joining to discuss Universal Quantum, and thank you for listening. If you enjoyed the show, please subscribe to Protiviti’s The Post-Quantum World and leave a review to help others find us. Be sure to follow me on all socials @KonstantHacker. You’ll find links there to what we’re doing in Quantum Computing Services at Protiviti. You can also DM me questions or suggestions for what you’d like to hear on the show. For more on our quantum services, check out Protiviti.com, or follow Protiviti Tech on Twitter and LinkedIn. Until next time, be kind, and stay quantum-curious.