Quantum computing currently costs $10,000 a qubit: just one of the reasons why it’s hard and expensive to scale. SEEQC is taking a different approach to building scalable million-qubit machines that can actually deliver on the promise of quantum computers and revolutionize computing.
Will they succeed?
Hard to say, but SEEQC just announced that they are building a commercially-scalable, application-specific quantum computer for pharmaceutical drug development. Merck has bought one, and the company is partnering with Riverlane and Oxford Instruments to make it a reality.
Support TechFirst: Become a $SMRT stakeholder
The real challenge here is to take quantum computing from a V2 rocket era, CEO John Levy says, and bring it into the SpaceX era. In other words, to generate real, massive, lasting, and provable value from quantum computers … which, let’s be honest, we haven’t really seen yet.
Here’s my story at Forbes … keep scrolling to watch the interview, subscribe to TechFirst, and get a full transcript of our conversation.
Watch: scalable quantum computing
Subscribe to TechFirst: scalable quantum computing
Transcript: SEEQC’s new approach to quantum computing
(This transcript has been lightly edited for length and clarity.)
John Koetsier: Quantum computing is having a moment. China recently announced a quantum computer that they said is 10 million times faster than the fastest supercomputers available today. There’s some doubt about that and not all details were published, but immediately after, IBM announced the world’s largest superconducting quantum computer at 127 qubits, and they hope to have a thousand qubit machine in 2023. Still not a hundred percent clear what quantum computers can do in the real world and what massive advantage they can actually deliver.
Quantum computing company SEEQC, however, just announced that they are building a commercially-scalable, application-specific quantum computer. It’ll be used for pharmaceutical drug development; it was bought by Merck.
And to learn more, we’re chatting with SEEQC CEO, John Levy. Welcome!
John Levy: Hey, thanks. Thanks for inviting me, John.
John Koetsier: Hey, super pleased to have you today. What are you building?
John Levy: Yeah, so, it’s interesting to think about quantum computing and where we are, because it seems like almost every day now we’re getting really interesting announcements on the hardware side, on the software side, even on the theoretical side. And it’s clearly, you know, this is a really exciting time and you’re seeing all these advancements.
But what we’re focused on really, is figuring out a strategy for how to scale a quantum computer.
So, I almost liken where we are now to either the early days of classical computing, say when the ENIAC was built, or even closer in to the early days of the internet when people were showing early examples of what could be but weren’t doing it on platforms that ultimately were able to scale. And it took, you know, a lot of invention in the case of classical computing, it took the invention of the transistor or the integrated circuit, the microprocessor, RAM memory, and all of that, before you could actually scale a really, like an enterprise-grade classical computer.
And we think the same is true with quantum computing.
So these announcements are, I mean, coming out of China and IBM, are fantastic. They’re really great and showing, I think, the direction we need to go. But what we have built and are building — continuing to build — are the elements of building a chip-scale, all-digital quantum computer.
So, what I mean by that is … today, if you look at superconducting computers, and that’s, I think, where you’re seeing these announcements … we’re still dealing with analog signals that have to be converted to digital and gone back to analog, from cold temperature to ambient to cold. There’s a lot of latency. It’s very difficult to use microwaves to control qubits and all of this stuff is, if you saw what it looked like, it looks like you’re in a lab … and that’s really what it’s like.
So, but we’re used to working with computers, real computers that are all based on chips.
So our chips are based on something called Single Flux Quantum, which is different than CMOS — much faster, much quieter, way lower energy — and it enables us to build functional chips for a quantum computer doing things like readout and control, eventually error correction, co-processing, etc. all in the same dilution refrigerator, all at chip scale, and all of these chips we design and we manufacture in our foundry.
So we think that this is, you know, as much as IBM will get to a thousand or thousands of qubits, for example, the question is how do we get to ten thousand, a hundred thousand, a million qubits so that we actually scale computers, quantum computers, to the complexity of the problems that large companies care about … and that’s what we’re doing.
John Koetsier: That sounds amazing. It sounds like a totally different approach. I mean, I’ve seen quantum computers and, frankly, they look pretty cyberpunk. [Laughter & crosstalk]. They look really involved, really detailed.
It’s like 18th century aesthetics meets 22nd century technology.
Help me understand what you’re actually building, how that’s actually different. It sounds like you’re building chips that are kind of more normal, in a sense…
John Levy: Right.
John Koetsier: … similar to the chips that we have in our laptops right now, but function at quantum computing — in quantum computing ways. Is that correct?
John Levy: Yeah. I mean, so what we are building are — so if you think about it, you have the qubit layer, the quantum layer, if you will, and then on top of that, everything else that you’re working with in a quantum computer is really in the classical domain. Right?
And so, so the question is how do you make it co-exist in the same environment as the qubits, instead of having everything exist outside, say, of a dilution refrigerator, how could you build an environment where everything exists within the same environment? And that’s what we’re doing.
So think about chips — and by the way, chips that have to be connected to each other too — and we maintain superconductive connectivity and speed in doing that. Well, these are unique problems that have to be solved, but this is what you need to do to architect real, working computers at scale. You’ve got to solve these IO problems, connectivity problems. They may sound like pick and shovel, but nobody’s ever done it before. And so a lot of this work and these kinds of chips, even though they are of classical domain, if you will, they’re meant to operate, say, at 20 millikelvin and operate not at, say, two or three or four gigahertz the way the microprocessors in the computers that we all use operate.
But our circuits are timed to operate at like 30 and 40 gigahertz, and at three to five orders of magnitude lower power. Because, you know, power is heat. Heat destabilizes qubits at cold temperature. And so when you scale up to large numbers, you don’t want to introduce much heat, otherwise your qubits don’t work.
At small numbers of qubits, they are just fine. So that’s what we’re seeing today. But again, we’re focused on enterprise-grade quantum computing that’s going to require tens of thousands, hundreds of thousands, maybe millions of qubits or more.
John Koetsier: That’s kind of mind blowing. When we’ve been talking about 10, 20, 50, 100 qubit machines [chuckling]…
John Levy: Yeah right. Right.
John Koetsier: … and you’re talking about million qubits. This is pretty groundbreaking, because you’re fusing this high-performance computer with a quantum computer in the same network infrastructure.
John Levy: Mm-hmm.
John Koetsier: Go into a little more depth on how that actually functions, because hasn’t it been in the past with a quantum computer that you create, basically a program. You send those instructions — almost to another dimension — to the quantum computer. Stuff happens. Something comes back out. You interpret that, and then you communicate back and forth again.
John Levy: Yeah.
John Koetsier: How is yours different?
John Levy: Okay, so there are two dimensions to this issue. So one of it is the, think about the guts of a quantum computer and signals going back and forth. So, that’s things like readout, and control, and potentially error correction, things like that. The other part that you’re talking about is the actual operations of an algorithm or an application, where, for example, let’s just take an example, you have a quantum front end of a problem, you’re trying to generate a new candidate for some theoretical backend where you actually have a working classical model.
And then what you want to do is you want to keep that process going and iterate. So, how would you do it today?
You try to generate the front end quantum candidate in a quantum computer. Get an answer out. Go out to the cloud. Put that answer in, that new candidate into your existing classical algorithm or application. Run that, see how that goes. Go back into your quantum computer … and keep going.
John Koetsier: Mm-hmm, mm-hmm.
John Levy: Well, we’ve said, ‘That seems kind of nuts. Why not actually build in real processing power and fast processing power, classical processing power, in the same quantum computer?’ So we’re using classical technology to do two things.
One is, for the actual operation of the quantum computer, just as a quantum computer.
And the second, is to think about having a very fast Single Flux Quantum coprocessor in the same quantum environment, so that you can actually iterate within the same environment. You don’t have to go outside to the cloud. And so we think that that architecture, again, lends, or leads to scaling at much higher levels … and at much faster speed, by the way.
John Koetsier: What will this enable? I know you’re working with Merck. You’re working on biotech and drugs and pharmaceuticals and stuff like that. Talk about what this is going to enable.
John Levy: Yeah. So, so look, our strategy is — so we have big vision architecture and we’re building all that. And on the other hand, the humility kicks in when you actually have to solve a real-world problem. And so what we’ve said is let’s try to narrow the scope of the problem to as small a thing as we can, and build the quantum computer to that level, co-designed with our applications partners.
So for example, the one you brought up is called the QPharma program. We’re leading it. Merck is our customer. But we brought in groups like Riverlane who have expertise not only on the operating system level with Deltaflow, but also in quantum chemistry. And so what we can do is we can reduce the problem set, co-design our quantum computer with Riverlane, so that we make sure that we are supporting exactly what they need and nothing more … and then deliver that — and that we’re doing that, by the way, in conjunction with Oxford Instruments and some other groups — and deliver that to Merck as a way of saying here is not just an example of how we can build a quantum computer, but it’s a purpose-built quantum computer that’s built on a scalable architecture, that we can scale to large numbers.
And so the idea is: start small, and as we scale, increase … increase complexity.
John Koetsier: What’s really interesting to me about that, is that the first classical computers were purpose-built computers …
John Levy: Totally.
John Koetsier: … computing firing solutions for the military. If I shoot something aiming in this direction, with this much power, with that angle of inclination, where will it go? [laughing] What will it hit, right?
John Levy: Yeah. At the very end of World War II, you know, and sort of we’re now in like the V-2 rocket or post V-2 rocket days. Right. The government was trying to figure out how do we track missile trajectories? And instead of saying, ‘Let’s build a general purpose computer and we’ll try to solve the problem,’ they said, ‘Let’s solve that problem.’
And in fact, it’s, you know, early days of Nvidia, it’s what they did when they were trying to port a specific game from a Sega platform over to an open source IBM or industry-standard IBM PC. The idea was, let’s not solve, let’s not build GPUs for the future, let’s solve this in this specific problem. And by doing that, we will eventually scale a system that will become a full-scale GPU.
John Koetsier: That is really fascinating, because, A, it’s humble. We’re gonna build something that’s going to solve this problem, not all problems.
John Levy: Right.
John Koetsier: B, it seems to have a higher chance of success because you’re building for a specific purpose, so that’s pretty cool. Everybody does want to know when the general purpose quantum computer is coming. [Laughter & crosstalk]. Everybody wants, when can I buy the box … that will tell me the answer, I can run any program on it, and I can get out good answers. Before we get to that question maybe…
John Levy: [Laughing] Good!
John Koetsier: How many massive problems are there that you think you can build custom, purpose-built quantum computers for? You’ve mentioned drugs already, drug design — that has a lot of stuff embedded there, biotech, other things like that. How many other types of problems are there out there that you’re looking at building custom, purpose-built quantum computers for?
John Levy: Yeah, so let me, let me kind of refine this idea about building custom computers. The vast majority of what we’re building is common among all the applications, right?
John Koetsier: Right.
John Levy: It’s just that because we have chip designers and we have a chip foundry, and we know how to co-design computers, we can mix and match various chips that we have, to some core algorithms, you know, QAOA or quantum variational eigensolver or something, whatever.
And we can actually begin to create families of chips that we can drop into a computer.
So, I want to call them instead of just custom built, they’re more … easily reconfigurable. And, by the way, that’s an interstitial strategy.
I think the core strategy, of course, is what you mentioned, which is to build fault-tolerant, error-corrected, general purpose, quantum computers.
That’s it. I think that there are, in almost every category, there are important, NP-hard, quantum problems. And one of the things that we focus on, my, our CPO, Matt Hutchings focuses on in particular, are making sure that the things that we’re working on are in fact amenable to a quantum solution. And not only that, but when we work on some of these things, these agreements with companies like Merck and with others, one of the first steps that we take is to benchmark what can be done best on a classical computer and at what level, so that we can actually say, ‘This is where we are today from a classical perspective. This is where we’ve benchmarked it,’ so that we can then measure the improvement that we make on a quantum basis.
And whether or not that’s in portfolio analysis, or whether it’s in a new catalyst for batteries, or, I mean, I was on the phone today. I was on a call, talking to a financial services company — a big one, it’s one that everybody would know — and where they think that quantum computing is going to be used for security and for identification and for, you know, with AI, for fraud detection and all the things that we think can be done and are being done on classical computers, but could be done in a much, much better way on a quantum computer, and particularly aligned with some AI work, and machine learning.
John Koetsier: I really like that, because one of the companies that I’ve talked to about quantum computing, you know, one of their go-to examples of how their quantum computer was being used was scheduling the paint booth for Toyota…
John Levy: Yep, I know who you’re talking about [laughing].
John Koetsier: … or GM or something like that. And I’m like, I understand the traveling salesman problem can be challenging for classical computers as well, and maybe this is sort of analogous to it, but that doesn’t seem like you’re solving world peace or world hunger or inventing the next, you know, antiviral …
John Levy: Yeah, but John. I mean, I, again, you know, being a little humble about all this and saying how do we narrow problem sets? I think we are at a stage where these kinds of things are really, really critical to show that they can be done, and that they have some commercial value.
The real question is, once you show that, what’s the strategy to scale? How do you scale that?
So yeah, the paint booth — I’m very familiar with the paint booth problem, I’ve heard about this actually from a number of companies. But, right. It’s not world peace, it’s not curing cancer, whatever. It’s not solving, or like I think about climate change. But it is a good starting place for building with a scalable architecture.
John Koetsier: Okay.
John Levy: If what we’re doing is building demos, if you will — and I think that’s kind of where we are, we’re like in the demo phase of quantum computing — and I think if that’s all we’re doing and that’s as far as we get, well, okay, so be it. But then we need to evolve to that next generation where we’re into the scaling idea. And again, you know, think about it, again, that’s what happened in building classical computers. Frankly, it happened to the internet.
John Koetsier: Yep.
John Levy: And if we think about the internet in the sort of seventies, eighties, and then you look at when things really took off in the mid to late nineties, and then there was a kind of retrenchment, because things had become open source, we needed broadband, we needed servers that weren’t custom built, that anybody could buy inexpensively. And before you knew it, suddenly all the tools and all the infrastructure existed for us to actually enter the real kind of internet era. [Crosstalk]. Before then, it was just demos.
John Koetsier: So let’s talk about that scaling then. You said earlier that we’re kind of at the V-2 rocket stage for quantum computing, and, if we’re really honest, not much changed in rocketry … even to the shuttle, which wasn’t really reusable, it was kind of reusable, the space shuttle … until we’ve had SpaceX, where you’ve actually got reusable, landing rockets.
So if we’re at V-2 stage right now for quantum computing, when will we get to the SpaceX stage?
John Levy: Well, I think, look, I think it’s all about having this kind of architecture that we’re talking about. We’re, we are on the third generation of our SFQ — of most of our SSQ circuits. Some of them were actually just, some of them are first-generation, but with these are chips that are, we’re trying to perfect and integrate and do all those sorts of things. This is not, you know, it is rocket science sorta…
John Koetsier: Haha, yes, or worse.
John Levy: … right? I mean, but, staying with your metaphor. But the fact is, is that this isn’t like some futuristic idea of wouldn’t it be great if? We’re doing this.
And by the way, there are one or two other companies in the quantum space — not in the superconductive realm, but more in like photonics, for example — that are taking a similar tack with building a chip-based system. I would argue that superconductive electronics offers a better solution for computing and for high speed computing than some of the other modalities, but forget that. Just the idea that of a scaling strategy, look, we know that in order to scale large complex systems, we have to reduce them to a series of chips. It just has to happen.
John Koetsier: Yep.
John Levy: The question simply is: What’s the methodology, what’s the technology, and where are you in the process? And we’re, you know, we are designing and manufacturing these things and making incremental improvements literally every week. I mean, one of the advantages of having a chip foundry across the street from where you do design and testing is that we are running wafers all the time. And, in fact, we hope to continue to expand our facility; we’d like to actually build a brand new facility to make our work even faster.
John Koetsier: So…
John Levy: ‘Cause this is what has to happen.
John Koetsier: So look into that crystal ball, and when do we get out of the V-2 era and when do we get into the quick and easily accessible, fairly inexpensive era?
John Levy: [Laughing] Well, I don’t know about the fairly inexpensive part, but I will tell you…
John Koetsier: Let’s say relatively inexpensive.
John Levy: … but yeah, I want to give you a metric though.
I mean, right now, the cost of building a qubit that you can — of the qubit with readout, control, tuneable couplers, things like that, right, is at about $10,000 per qubit, per system. Not long ago, it was around $40,000. So there’s been a real improvement there. But imagine for a moment that you could do all of that with, I don’t know, five or six chips, something like that, and where you could multiplex to a lot of qubits. Well, suddenly, your cost goes from the, let’s call it $10,000 range down to, you know, not many dollars to build a set of these sorts of chips that you can build on six- and eight-inch wafers, where you get a lot of these chips. So you can imagine that the cost would go down. Now…
John Koetsier: Yes.
John Levy: …is it going to be something on everybody’s desk? I’m not so sure about that. I’m not sure you need that, but that’s okay. But I do think that within the next three to five years, what you’re gonna see is you’re gonna see quantum computers built on this — on this and other scalable platforms. Again, I think you’ll see a photonics-based one, and you’ll see from us, you’re going to see a superconductive electronics one … superconductive qubit version.
John Koetsier: And that not many dollars. What’s that specifically? You’re at $10,000 a qubit right now. Are you talking $500? You’re talking hundreds?
John Levy: I think hundreds.
John Koetsier: Okay. Okay.
John Levy: And so, that’s an important development because, I mean, if, again, let’s just say for a moment, your goal is to build a million qubits, well, a million qubits times $10,000 per qubit is a [laughing] lot of money.
John Koetsier: Yes it is.
John Levy: You know, maybe Google can afford to do that, but not many others could. So we need to find ways not only to address issues of latency, and of heat dissipation, and speed, and noise, and RF interference, and all these other things that you need to solve, because all of them need to solve for a scalable solution, but from a cost and complexity perspective, you need to get it down operating at chip scale.
John Koetsier: Super interesting. John, thank you so much for taking this time and sharing what you’re doing, what you’re working on, and wish you the very best. Sounds like cool work.
John Levy: Thanks for inviting me. I look forward to following up and having you come and see what we’re building.
John Koetsier: That would be amazing. Talk soon.
John Levy: Thanks.
TechFirst is about smart matter … drones, AI, robots, and other cutting-edge tech
Made it all the way down here? Wow!
The TechFirst with John Koetsier podcast is about tech that is changing the world, including wearable tech, and innovators who are shaping the future. Guests include former Apple CEO John Scully. The head of Facebook gaming. Amazon’s head of robotics. GitHub’s CTO. Twitter’s chief information security officer, and much more. Scientists inventing smart contact lenses. Startup entrepreneurs. Google executives. Former Microsoft CTO Nathan Myhrvold. And much, much more.
Consider supporting TechFirst by becoming a $SMRT stakeholder, and subscribe on your podcast platform of choice: