Yuval Boger interviews Christian Weedbrook, founder and CEO of Xanadu. They discuss Xanadu’s photonic approach to gate-based quantum computing, the advantages of room-temperature operation, and the company’s plan to scale through optical networking. Christian also describes PennyLane, photonic error correction, customer engagements, and Xanadu’s target of building a 500-logical-qubit system by 2029–2030. The conversation also touches on partnerships, manufacturing, likely early applications in chemistry and materials, and much more.
Transcript
Yuval Boger: Christian, thank you for joining me today.
Christian Weedbrook: Thank you for having me. I appreciate it.
Yuval: So who are you and what do you do?
Christian: My name is Christian Weedbrook and I’m the founder and CEO at Xanadu and we’re building photonic quantum computers.
Yuval: How do photonic quantum computers work and how are they different from gate-based quantum computers?
Christian: Well, thanks. So, the photonic quantum computers are still gate-based, which is important, but it really just comes back to the idea that you can build a quantum computer out of either electrons or photons, really anything that you can kind of extract the quantumness, so to speak. So entanglement, superposition, interference. If you can do that, then you can build a quantum computer out of anything. We’ve chosen photonics because it has a number of benefits. Essentially you can compute at room temperature, which is a huge benefit for our approach. So no laser cooling, no cryogenics or anything when it comes to the qubits, the gates and the measurements. We need a little bit of cooling for the initialization stage of the quantum computer. We can also work with and do work with large-scale manufacturers when it comes to the foundries, which is great for when you want to do high-volume manufacturing and use the latest tools as well. And then also the way that us and pretty much everyone else scales up ultimately to a large-scale quantum computer is using light or networking, so fiber optics. And we’re the only company to have demonstrated how that works with a paper that was published in Nature last year. In fact, it was four server racks that we called Aurora, and they were networked together. We could have actually scaled up to hundreds of server racks if we wanted to, meaning that we know how to scale up using photons or light.
Yuval: How does one program a photonic quantum computer? Is that the same way that you would program an IBM or a QuEra machine?
Christian: Yeah, very much the same way. You know, at the high level, we’re still dealing with qubits and gates, and they have parameters. And so at that high level, it’s very similar. And in fact, for us, we created PennyLane, which is one of the most widely used software offerings in the quantum industry. It not only works on our hardware, but other folks’ hardware as well. And the reason I bring that up is because that’s the way you ultimately will program our quantum computers. And ultimately it’s the same way when you abstract away how the hardware is built.
Yuval: You know, when I have CEOs on this podcast, they usually tell me how great their approach is and the question that’s obvious is, well, if it’s so great, why isn’t everyone doing it?
Christian: It’s a good question. And what we try to do is we try to be as open about it as possible. And we think that there’ll be a few winners, not too many, but a few, maybe a handful. We obviously think we’ll be one of the winners too, but there’ll be others. The market size will be astronomical in the future. And there’s also a lot of great approaches and a lot of smart people working on them. So it’s really hard to tell. We have high confidence in our approach due to the photonic-based approach, but we don’t think we’ll be the only winner.
Yuval: What can people do today with your computer?
Christian: So we’re fully focused now on building a large-scale quantum computer. So hundreds of quantum computers networked together, and we’re fully focused on that. And so what that means is we’re not making available any more of our quantum computers on the cloud. That may change over the coming years, we’ll see, but focus is very key for us. In the past though, we’ve had our X8 machine online that allowed people to sort of play around with a photonic quantum computer. And then perhaps more famously, we had our Borealis quantum computer that demonstrated quantum supremacy, at the time the largest demonstration. And it was the only at that point in time and the first quantum supremacy quantum computer to be available online. In fact, online with our Xanadu cloud and also available on AWS, and it allowed people to play around with a computer that you could never hope to simulate the output using a conventional computer in any reasonable amount of time. So those were the ones that we had available. And as mentioned, we’re really focused heads down building this large-scale quantum computer that we define with, you know, there’s many, many parameters, but at a high level, up to 500 logical qubits.
Yuval: Is error correction done the same way? Can you tell me a little bit about what codes you use or what error rate you expect to be or what are the challenges in doing quantum error correction on a photonic computer?
Christian: Yeah, at a high level we all have errors that are caused by whatever approach we’re using, it kind of comes out depending on the physical system of choice. For us, we use photons and photonic loss is the reason for our error. So we really need to combat loss and it’s very similar to other approaches, you choose a type of error correction code. We can ultimately leave that choice to the end because we have such flexibility in our architecture, but we do spend a lot of time focusing from a theoretical point of view on LDPC codes, just because of the ultimately low overhead that’s needed in terms of extra hardware or redundancy. So they’re really great for that side of things. For us, we can also use LDPC codes. Not everyone can because there’s some restrictions there based on the hardware, which we don’t have, which is really cool.
The four server racks I mentioned that were networked together, we called Aurora. It actually has everything in place for error correction and fault tolerance. All the mechanisms and components are there. We actually showed real-time error detection and correction for the first time for photonic systems, also published in Nature. What we need to do and continue over the next couple of years is to reduce loss, that mechanism that’s causing errors. And one of the big ways to do that is through not only error correction, but also improving the physical performance of our chips as well. And you can do that by working more closely with foundries, which is what we’ve been doing and what we intend to do more of.
Yuval: There are a couple of other photonic vendors. Some have even raised more money than I think you’re intending to raise. How do you differentiate your approach versus theirs?
Christian: Yeah, a lot of great competitors in both electronics and also, as you mentioned, photonics. I would say not all photonics approaches, even though they’re photonics, are created equally, the same in the electronic-based approach, where there’s differences there. The great thing about our approach is, we need far less cooling than the other photonic companies, by far. And it seems to our knowledge that the others need cooling throughout every aspect of their quantum computer, whereas we don’t need it through the qubits and the gates and the measurements. So savings in energy, savings in iteration time, savings in making new types of cryogenics and that. We have huge savings there. And also we haven’t needed to raise as much money as others because we don’t need as much cooling and therefore we’re much more cost effective as well. So I would say we’re far ahead in terms of other approaches, in terms of the software side. PennyLane has been around since 2018 and we’ve built a loyal and dedicated user base of developers as well.
Yuval: You mentioned PennyLane. How did that start? I mean, PennyLane has been out much before you’ve shown a computer. For years, I believe, Xanadu was known more for PennyLane than for the hardware that you’re building.
Christian: Yeah, we think that’s true, actually. I think that was a good thing. It was good to release something from Xanadu while we were building out the hardware. And the easiest thing to release was the software, which we did. PennyLane actually started off as– today it’s full stack, meaning it goes end to end, including compilation, which we call Catalyst. One of the most widely used, if not the most widely used compilation stack in the industry, all the way down to controlling the hardware, and at the top layer, how people can use it in an agnostic way. But it actually started off as a quantum machine learning package. And that’s why you see that it’s rare in the sense that it has a differentiable nature to it, where you can optimize throughout the whole stack as well. And that’s because it started off as a QML sort of open source package, but it really developed from there. We’ve had, you know, PennyLane is available in over 150 universities as part of the teaching curriculum worldwide in over 30 countries. It’s also used by folks like AMD, Lockheed Martin and many of our partners, and the way to sort of attract really great partners that we have and ultimately customers as well. So we’re really proud of PennyLane and we continue to listen to users and try to make it as awesome as possible.
Yuval: You mentioned 500 logical qubits as sort of a target. When you think about that computer or a set of networked racks that reach 500 logical qubits, what is your estimate for how large, how physically large that system is? How much power will it consume? And to the extent that you’re able to share, how much will it cost to make it?
Christian: Yeah, all good questions. Our footprint, we believe, is one of the lowest in the industry. Around three to five tennis court sizes is what we’re anticipating. And the types of problems, it’ll be anything you can solve up to 500 logical qubits. The energy cost, we estimate at this point in time to be relatively on order of magnitude the same as what it costs for the same amount of server racks for a traditional data center. So perhaps something like single digits megawatts. And the cost, we’ve been public about that. The first one will be the most expensive, like in any technology, the first product, and reduced over time. It’ll be about a billion dollars to build out something that’s up to 500 logical qubits.
Yuval: When will that machine be available in your estimate?
Christian: Yeah, we’ve been very public about this. We’re aiming for 2029, 2030 for that.
Yuval: And you’re raising now a good amount of money, I believe, through a SPAC transaction. If I were a potential investor, why should I invest in Xanadu?
Christian: Two major points that when we talk to investors are the following. The first is, we’ll be the first pure play photonic quantum computing company to go public. So investors are unsure who the winners will be. And so they’re making bets on different modalities or hardware. They can now invest in photonics for the first time. The second one is where the big sort of, you know, aspect of Xanadu is that we created PennyLane, one of the most widely used software offerings anywhere in the industry. And now investors have the chance to invest in PennyLane through Xanadu. And so it’s really those two things, photonics and the PennyLane product.
Yuval: Are you able to monetize PennyLane?
Christian: Well, I would say that historically, up until this day, it’s all been about ecosystem building and having lower barriers to entry to build that ecosystem, which the team has done a fantastic job. At some point, I’m sure we’ll think about how to monetize PennyLane, and that really follows the well-worn playbook of open source software, whether it’s quantum or not. So offering tiered versions or tiered pricing from a free level to an enterprise level, and then offering support and services as well. So we’ll definitely consider that.
Yuval: We spoke about PennyLane, and I see behind you an Abbey Road Beatles record. What’s the Beatles story behind Xanadu?
Christian: Oh, I just personally love the Beatles. They’ve been with me now since my teenage years and haven’t left. And, you know, I love the Beatles. They’re very inspirational to me. I just get a lot of joy out of it. So that’s kind of the origin there.
Yuval: And in terms of your personal story, how did you get started with this?
Christian: Well, like a lot of people, I came from academia. So I started in Australia where I’m originally from, I did a PhD in photonic quantum computing. And then after that was a postdoc at MIT and then a postdoc at the University of Toronto. So really came from that heritage of academia and contributing to photonic quantum computing research. But when I was younger and still to this day, I’d read a lot about entrepreneurs and businesses, a lot of nonfiction, a lot of biography. So those two things really have come together. And so I love what I do and I love how we’re all working towards this big scale problem.
Yuval: I believe you announced several partnerships recently. Which ones are you most excited about?
Christian: Well, all of them are very exciting. We have a great set of partners and customers. The two latest are very exciting as well, including AMD. We announced that on Monday. We also announced Lockheed Martin a few weeks ago, and also partnerships with governments ’cause the defense plays a big role. We announced up to $400 million partnership of matching funding from the Canadian government, both federal and provincial. So we have great support from Canada. We’re proudly Canadian and looking to, you know, make our mark on the world stage.
Yuval: You’ve been in this for quite some time. What have you learned about the quantum world in the past, say, 12 months?
Christian: Well, I would say from a commercial point of view that it’s getting very close to now, this large-scale quantum computer. For us, what I can say is that our engineers and researchers, we see the end in sight now. You know, a lot of the major challenges have been solved over the last 12 to 24 months by the team. So, you know, it’s all about the hardware. And from our point of view, we can really see us getting there. Whereas when we first started, it’s like, wow, people are giving us money to do this. That’s kind of cool. Let’s see how far we can go. And I think that’s really changed our perspective over the years. And I think that’s been validated through peer review research in Nature, through government funding, both in the Canadian, US, and also by our partnerships as well.
Yuval: When you work with customers, beyond sort of the fundamental technology and the scaling path, how does the project work? Do you just say, here’s our environment and go knock yourself out and you can run it on the computer? Is there a very extensive co-design process together with the customer? How do these engagements work?
Christian: I would say it’s par for the course for the whole industry, how we make money. And it’s really a package deal. So we offer cloud access, so access to our hardware. We offer access to our PennyLane software and access to our use case algorithms team as well. And so it’s a package deal where the customers pay money for that. And out of that comes understanding of where quantum computing will really help one day, patents, co-patenting, individual patents come out of it as well. So it’s really that mold. We’ve also sold cloud time, just cloud time on our hardware to government labs in the past. But this is the sense. And ultimately the final version, there’s a couple of ways to make money from this. One of the big ones initially is building our own data center and offering cloud computing as a service to customers to access over the cloud. So that’s really, you know, we’re setting ourselves up now by working with these large partners.
Yuval: As popular as cloud access is, it seems like some customers really want the computer to be on premises. Is that something that you’re offering or intend to be offering?
Christian: Intend to be offering for sure. Once we build this large-scale quantum computer, hundreds of our own server racks populating it, three to five tennis court size roughly, we’ll be also intending to sell these server racks to others, just because like you said, data sovereignty is becoming more and more common in the industry. And so we’ll have that option as well. So cloud access and also selling individually so people can have it on-prem.
Yuval: How and where is it manufactured?
Christian: Well, I would say a lot of it is done in-house, but the big one that comes to mind that’s not done in-house is the foundry work. So we design all our chips and patent them ourselves, but then work with a variety of foundries around the world, in the US, Taiwan, Europe, and Singapore. And they make or fabricate the chip designs and then send them to us. And after that we take over pretty much everything else, you know, the packaging, test and measurement, packaging and the final assembly as well.
Yuval: Some people say that it’s unlikely for a company to be truly full stack, to design the hardware and software and then deliver it and manufacture and support and program. What do you need from the rest of the quantum ecosystem to make Xanadu even more successful?
Christian: That’s a good point. I think inherently we don’t want to do everything ourselves. It would be nice, but at the end of the day, no one cares about your stuff more than yourself. And so you have to move faster. I love the idea that hardware and software folks are in the same office so they can work side by side and iterate much faster. So there’s just so many benefits by doing it essentially yourself. Now we can’t do the foundry work, so that’s a good example of the quantum ecosystem, working with those foundries. There’s other parts, we’ll see what other companies are out there as they get formed. There’s new quantum computing companies getting formed all the time, offering different things. We keep a close eye on the quantum industry to see if there’s overlap. We also are interested in working with others if they’re interested in buying our chips to help them with interconnects and stuff. So yeah, that’s really how we see it at the moment and always keeping our eyes open.
Yuval: As we get closer to the end of our conversation, I’m curious about applications. Do you see a particular application, chemistry or optimization or machine learning or something else, that would deliver value perhaps sooner than other applications to the end users?
Christian: Yeah, basically, you know, what can you do with many parameters, but what can you do with, say, a bunch, hundreds of logical qubits? And honestly, to be fair, the rest of the industry, I think we’re all coalescing around quantum chemistry and material design. That’s probably the lowest hanging fruit in terms of the requirements. So next generation batteries, next generation solar cells, discovering and synthesizing new materials. That’s probably the earliest application space for us and others in the industry.
Yuval: Have you filed a petition to change Canada to a Q, the C to a Q?
Christian: We’re working on it. We’ll let you know how that goes.
Yuval: Very good. And last, a hypothetical. If you could have dinner with one of the quantum greats, dead or alive, who would that be?
Christian: Oh, well, I would have to say probably Einstein. He contributed a lot to quantum and was also a skeptic to the end. But he would be fun to sort of have dinner with.
Yuval: Christian, thank you so much for joining me today.
Christian: Thanks for having me. I appreciate it.
—
Yuval Boger is the Chief Commercial Officer of QuEra Computing.