IEEE Quantum Podcast Series: Episode 11
A Conversation with Pete Shadbolt
Co-Founder and Chief Strategy Officer, PsiQuantum
Listen to episode 11 (MP3, 27 MB)
Part of the IEEE Quantum Podcast Series
Brian Walker: The IEEE Quantum Podcast series aims to inform on the landscape of the quantum ecosystem and serve as the leading community for all projects and activities on quantum technologies. This episode features Pete Shadbolt, co-founder and chief scientific officer of PsiQuantum, a Palo Alto based startup building the world's largest silicon photonic quantum computer. Pete discusses the challenges and likely benefits of achieving a million-qubit fault tolerant quantum computing system. Pete also shares his insights on why students and young professionals might want to further explore this growing field.
Pete, thank you for taking time to contribute to the IEEE Quantum Podcast series. To start, can you give our listeners a brief summary of your background leading up to your current position?
Pete Shadbolt: Yeah, sure. I’m Pete Shadbolt, I'm the co-founder of PsiQuantum. So I've been working on quantum computing for about ten years. Prior to starting the company, I worked in academia initially as an experimentalist. So I spent about five years at the University of Bristol in the UK building and testing small photonic quantum processors. I went from Bristol to Imperial College where I worked with one of my co-founders, Terry, on the sort of math and theory of fault tolerant optical quantum computing. And then I moved out here to California five years ago now to start this company.
Brian Walker: So, Pete, how did you first become involved with IEEE
Pete Shadbolt: Yeah. So I spoke at IEEE Rebooting Computing back in 2019 in San Mateo. And I think what's interesting about PsiQuantum is that, our thesis is that in order to build a useful quantum computer, we're going to have to leverage the incredibly powerful tools of the semiconductor industry, chip manufacturing, basically. And so relative to some quantum computing that some quantum computing companies that are more close to, you know, physics research or ion trapping or something like that, we do a lot of semiconductor manufacturing work. And that puts us in a position of, you know, I hope being interesting to people from IEEE. Really enthusiastic to engage with you guys.
Brian Walker: So what are your views on the current state of the quantum computing space, particularly in comparison to your company's approach?
Pete Shadbolt: So quantum computing is an idea that is decades old now, and there are all sorts of large and small teams of people working very hard to bring this technology to life. I think the biggest shift and one of the main aspects of differentiation for PsiQuantum is around whether you're building a big quantum computer or a small quantum computer. And what I mean by that is that five years ago, I think it's fair to say this sort of default position, if you were starting a quantum computing company or pursuing this technology, was to take the demo systems that we all had in our university research labs, scale them up by about an order of magnitude to get into the regime of about 100 qubits and hope that we would find commercially valuable applications with those systems. That's what's referred to as NISQ or noisy intermediate scale quantum computing. And it was a very sensible idea five years ago to try that approach, 100 qubits, even without error correction. That's a system that's very, very hard, if not impossible to simulate on a conventional supercomputer. And so the reasoning that, you know, hopefully there's something we can do that's useful with a system like that was certainly defensible. PsiQuantum on the other hand, took a very different approach when we started the company, which was to exclusively target a fault tolerant machine that could run error correction and thus continuously suppress error in the machine. And that's something that, you know, even a decade two decades ago, people thought would be necessary. And we kind of staked our bet on the premise that it would be necessary and that you would need at least a million qubits to deliver on the promise of quantum computing. That's, of course, a much, much larger system than the, you know, 50, 100 qubit NISQ systems that we're seeing today. And so, you asked about kind of the state of the space. I think, you know, over the last five years, a lot of time and money has been spent on this. People have made extraordinary scientific demonstrations of small scale systems. And a lot of time has been put into trying to find algorithms that can do something useful despite the errors and the noise that you see in these non-accredited systems. And unfortunately for the world, nobody has found anything. So today we do not know of any commercially valuable applications that can run without error correction. As far as the world knows, you still need a million qubits to do anything useful. And that's sad because it would have been a nice shortcut. But it's good for
PsiQuantum, because we've spent all of our time and money over the last five years on a million-qubit machine and building the technologies that we need to actually deliver on the system at that scale, most notably the semiconductor manufacturing piece. So I think we're seeing the whole world transition to a recognition that you're going to need a large scale error corrected machine. And in fact, Jeremy and myself were at the White House a couple of months ago where we saw many of the big players in the quantum computing space openly acknowledging that you're going to need a system of that kind of scale to deliver on the promise of this technology. So I think that's a really interesting, encouraging, healthy shift in the industry.
Brian Walker: So Pete, what are some of the initial challenges you face in trying to build a million qubit system?
Pete Shadbolt: Once you recognize that you're going to need a million qubits, the type of technical work that you engage in becomes quite different. Know if you're just trying to hack together a 50-qubit system, you can do that using the kind of techniques that you would associate with a university research group. If you're trying to build a million-qubit system, that's going to be, you know, building scale, very high-power cryogenics, huge numbers of chips, you get into a completely different regime that looks much more like building a conventional supercomputer. And where your challenges are no longer quantum, the challenges that you're trying to solve are really more to do with cooling, power, control electronics, connectivity, manufacturability, even not quantum problems. And we're starting to see increasingly, you know, serious investment and attention from the big players on solving those problems, which I think, again, it's just healthy and positive for the whole industry.
Brian Walker: Can you give a little more detailed information on the engineering challenges faced in building a quantum computing system of this size?
Pete Shadbolt: As far as the engineering challenges that you encounter once you set that as your target? I already touched on them and maybe I can expand a little bit. So, the first piece is manufacturability, where you need some way to manufacture millions to billions of components with very high precision, very good yields, very good integration, very good performance. And our perspective from the beginning has been that there are really only three institutions on the planet who can do that sort of thing. And they are TSMC, GlobalFoundries and Samsung. Intel, of course, is a special case. They got out of the foundry business. They're now getting back in. I should probably add them to that list. But these tier one semiconductor manufacturers are really unparalleled in their ability to manufacture huge numbers of components that will work. And it's been our conviction for a long time that you basically have to build your chips, build your qubits in those extremely mature manufacturing processes if you want to have any hope of yielding millions of qubits that are actually going to work. Now, the challenge with that is that TSMC, GlobalFoundries, etc. are not in the business of doing crazy science fiction stuff like quantum computing, you know, they're building laptops and cell phones and they are pretty supply limited in terms of the capacity that they have, especially right now where we're seeing automotive manufacturing lines shutting down because they can't get the chips. And so it's really pretty, pretty extraordinary as far as I'm concerned, that we have gotten into that regime over the last couple of years. We've put six tools into the production line at GlobalFoundries. We're building thousands of wafers worth of silicon, shoulder to shoulder with laptops and cell phones. We've introduced a brand-new superconducting material into that manufacturing process that's allowing us to build, you know, very large numbers of quantum devices in that context. And that's been a huge amount of the work that we've done over the last couple of years. And then beyond the manufacturing, there are things like cooling power that are currently constraining all approaches to quantum computing. If you need to cool things to millikelvin temperatures, it's very hard to get more than micro watts of cooling capacity, which is very, very limiting. Again, when you think about a million-qubit machine with photons, we are fortunate that of course the photons themselves don't feel heat. That allows us to run a higher operating temperature where we get thousands of times more cooling capacity today and where we ultimately expect to be able to access millions of times more cooling capacity than you can get at millikelvin temperatures. So the fact that the photon is indifferent to temperature doesn't feel electromagnetic interference, that's pretty advantageous. We also have advantages in terms of connectivity. So far, people have been able to show fantastic demonstrations using individual chips. But of course, you're not going to fit a million qubits onto a single chip. You're going to have to connect chips together. And so the fact that we can just send single photons into optical fiber, the same optical fiber that you find in a data center. And we've already been able to demonstrate teleportation and entanglement between chips using that regular optical fiber. That's a huge deal as far as, you know, scaling out and building much larger multi chip systems and something that's quite difficult to achieve in competing approaches where you typically have to convert from quantum information stored in a massive qubit to a quantum information stored in a photon, send that photon optical fiber and then convert back on the other side. That conversion or transduction is technically very challenging, even though it's not forbidden by the laws of physics. And then the final thing actually that's worth touching on is the control electronics. So it seems maybe, like an aside to the main engineering problem. In fact, it's absolutely, crucial and very limiting for scaling up current systems. We have been working very hard on that and we now have a control chip with about three quarters of a billion transistors that is built in a conventional 22 SGX processor, GlobalFoundries Dresden. And we're able to run that chip at cryogenic temperature. So that gives us a much more sophisticated, much more capable control where we can put that control very close to the qubits themselves. And that's going to be really enabling for building bigger and bigger systems in the future.
Brian Walker: So, Pete, where do you see quantum computing making notable impacts? What applications do you think are ripe for quantum-based solutions?
Pete Shadbolt: It's worth saying, of course, it's absolutely possible that somebody will come through with a breakthrough in this algorithm in the next few years. And I hope that they do because it will make life easier for everyone. But we, along with many of the big players, are basically operating under the assumption that that's not going to happen and we're going to need a big system. And then when it comes to applications for those big systems, there are kind of a handful of categories that we can put the nine useful quantum algorithms into. Codebreaking is probably the best known. So factoring large semi primes to break RSA encryption and similar, we don't think that's commercially particularly interesting. And it also requires a pretty big quantum computer. You will have heard a lot about optimization problems. As far as we know, optimization is interesting on very large quantum computers, but doesn't seem to give you profound speed ups. And really, as far as we're concerned, the most interesting quantum algorithms for a first system pertain to simulating things and specifically simulating quantum mechanics, simulating molecules. And so that brings us to the engagement that we have with our customers. We're very lucky to have a list of customers and partners from, you know, Fortune 500 type companies. Interestingly, these users have become increasingly sophisticated over the last five years or so. And big banks, car companies, pharmaceutical companies, they are hiring PhDs in quantum information. They are teaching themselves about how to program quantum computers. And they're getting increasingly serious about preparing for the existence of this technology. And yeah, we are now working with a large bank, five large pharmaceutical companies, a large car company. The list goes on basically to write these algorithms and evaluate them in terms of the resources that we would be required to actually get a significant quantum advantage for these applications. And the majority of those are in things like materials design, drug design designing and the catalysts for lithium-ion batteries. So basically, simulating molecules, simulating reaction chemistry, that type of thing. And that's again, where we see the most significant early advantages for quantum computing.
Brian Walker: Do you have any advice or insights that you can share with students or young professionals who might be interested in a quantum career?
Pete Shadbolt: I spoke to quite a lot of young people who are maybe software engineers, maybe they're semiconductor people, maybe they are from some field other than quantum computing, and they want to know whether you need a quantum background to come and work with us. And I always make the point that the answer is no. The majority of our engineering team have no quantum background at all. And that's because the technology is reaching a level of maturity where most of the problems to solve are engineering problems as opposed to quantum physics problems. And so we're extremely lucky to have such a, you know, experienced, powerful team of people who, again, don't necessarily have to have any quantum background on the quantum side. There is also, you know, huge opportunity. And, you know, we fight tooth and nail to get the best people, whether it's in quantum algorithms, quantum tolerance, error correction, etc.. That's a deeply mathematical regime that absolutely requires, you know, an extensive academic education to do that type of work. And it's an increasingly sought-after skill set. I think what's interesting there is that there are a large number of people around the world working on these algorithms, very smart people, and increasingly there are opportunities for people to instead work on fault tolerant algorithms. And that's quite a different type of work. So writing code, designing the algorithms that are going to run on a fault tolerant system involves a completely different set of tradeoffs to what you might be used to if you've been designing algorithms for machines. And I think, you know, hopefully it's a kind of interesting opportunity for those people. And if they're interested in working on such things, they should get in touch with us.
Brian Walker: Pete, thanks for your time today. Do you have any final thoughts you'd like to share with our listening audience?
Pete Shadbolt: Oh, I really appreciate you taking the time and inviting me to speak. I I've already said that it's a huge privilege for me to get to work on such an exciting technology. And I want to give a shout out to a large number of incredibly hard-working people, talented people who actually make this all move along. And then, yeah, I've already said this, but I'll reemphasize I think quantum computing is a really healthy stage where we're seeing customers and governments, you know, seriously engaged on the applications side, increasingly enthusiastic about not just kind of dabbling in applications, but really doing their homework on fault tolerant resource counting, really calculating with precision, you know, how many gates, how many qubits you going to require to do something that's genuinely valuable? I think that's a really, you know, new and mature tone for the applications work that certainly that we do with our customers. And then on the hardware side, it's also just extremely exciting to be, you know, embedded in an industry that is waking up to the fact that it's going to have to build a building scale machine and really getting serious about the very hard, very expensive and totally solvable engineering problems associated with doing that. So, yeah, thank you so much for having me on the call and I look forward to lots more interactions IEEE.
Brian Walker: Thank you for listening to our interview with Pete Shadbolt to learn more about IEEE Quantum. Please visit our web portal at quantum.ieee.org.