Julien Camirand Lemire (Nord Quantique): One Cavity, One Error-Corrected Qubit

Julien Camirand Lemire, co-founder and CEO of Nord Quantique, is interviewed by Yuval Boger. Julien discusses Nord Quantique’s bosonic code approach, which uses microwave photons in superconducting cavities to achieve quantum error correction without large qubit overhead, including their GKP and Tesseract code demonstrations and a quantum error correction gain that has more than doubled since 2024. They explore the company’s roadmap toward 10^-7 to 10^-9 error rates by 2032, the four-qubit system entering operation this year, and how Quebec’s Sherbrooke ecosystem has enabled a capital-efficient build. They also discuss high-rate codes, transmon-based scaling challenges, and much more.

Transcript

Yuval: Hello, Julien, thank you for being here.

Julien: Hi, Yuval, it’s a great pleasure to be here with you.

Yuval: So who are you and what do you do?

Julien: So I’m Julien Camirand Lemire, I’m co-founder and CEO of Nord Quantique. W e are a quantum computing hardware company building differentiated technology that really is anchored into error correction for scaling purposes.

Yuval: And you’re in Canada, right? That’s Canada with a Q, I believe.

Julien: Yes, Canada with a Q. We’re in Quebec, Canada, so it’s even more Q in the name. But yeah, we’re based in Sherbrooke, so we have offices here in the Sherbrooke ecosystem that is a pretty strong powerhouse in quantum in Canada, but also have offices in Montreal, which is well-known.

Yuval: So you got me interested in differentiated technology. Please explain.

Julien: Great. Yeah, I think, so we are a spin-off of this Sherbrooke ecosystem, as I’ve been mentioning. So we incorporated the company back in 2020. And at that time, I think it was pretty clear what we wanted to do, which is really build a technology. So a type of qubit that is capable of running quantum error correction. And the reason we were doing this is really looking at what the industry has been doing for the past 10 years. And I think we should just appreciate all the progress we have done as an industry in building quantum computing hardware that is more complex and scaling as well. So meaning that we have now as an industry systems that’s up somewhere between a hundred and thousands of qubits that are becoming more and more available. But despite this great progress, I think there are still some challenges, some very hard challenges that the industry is facing. And one of these challenges is errors as we are aware. So these qubits, they make errors, they’re prone to noise, and we need to deal with this. And the best way that the industry has started to also acknowledge is the need for quantum error correction. So this sort of error management layer that sits into the quantum computing stack that is there to detect and correct errors as they occur during a quantum computation. And the challenge with quantum error correction usually is the fact that it consumes a lot of resources. So you need some form of overhead inside your quantum computers to deal with these errors. So this overhead often comes in the forms of number of qubits. So you need a certain number of qubits more in your quantum computer to get there. And usually this factor is somewhere between 100 to 1,000. So that requires a larger and larger size system. But nothing tells you that this redundancy that is needed by quantum error correction has to come in the form of more qubits. And this is where we approach things a little bit differently at Nord Quantique. So the thing that you need for quantum error correction is a form of redundancy. And what we’re building is a technology that is called bosonic codes. These bosonic codes are made of microwave photons trapped into superconducting resonators. And the great thing about them is that you can trap multiple microwave photons inside a single cavity or single resonator, which provides you with the redundancy you need to do quantum error correction in a single cavity, which is really the embodiment of the hardware we’re building at Nord Quantique. So yeah, this is really what is differentiated about this technology is how we deal with quantum error correction, really trying to lower the overhead, sorry, to achieve quantum error correction in quantum hardware.

Yuval: There are other approaches, right? Amazon is developing a chip. Alice and Bob has a chip that they say will reduce the number of qubits required because I think it only, you know, it de-amplifies one type of error and you have to deal only with the other. How is that different than what you guys are doing?

Julien: Yeah, definitely. So bosonic codes is a large space. I think this just from the quantum perspective, so this is why we like these codes, but also just as the types of qubits we can build with bosonic codes, there’s many of them. Alluding to Alice and Bob who is developing certain types of cat qubit that is, as you’re mentioning, reducing one error channel. In our case, we use a different version that we call grid states. So the most notable example of these grid states is the GKP qubit. So we did some demonstration of that qubit early 2024. I think we’ve released our demonstration there. In these qubits, you can do full quantum error correction on a single of these qubits, so correcting for both X and Z errors, and this is what you will find in our 2024 paper. Beyond this also, these codes can be expanded to what we call multimode codes, so multiple, like in these cavity you can store multiple photons, you can also store photons of different frequencies, and these enables you to run different types of code, like the Tesseract code, which is a different example of the grid state that we have released in 2025.

Yuval: Do these cavities require cooling to near absolute zero, or do you operate at room temperature?

Julien: Yeah, thank you. Yeah, so we’re based on a superconducting platform, so what we use are the same building blocks as any superconducting qubit hardware, so we use resonators, of course. We use Josephson junctions to couple these resonators together and also to control them. So these operate at low temperature. But I think what is differentiated also from our architecture is that we need a smaller cryogenic environment to run systems in the end. So because we save resources for quantum error correction, there’s some layer of scaling that we don’t anticipate having to do. I mean, we’re reducing the overhead for quantum error correction by a large fraction with this cavity. And the purpose is really to build quantum computers, yes, in cryogenic environment, but in a constrained size. So something that will fit into 10 meters squared for more than 2000 of these logical qubits.

Yuval: So to get to real application, one doesn’t need a lot of qubits, just needs a large enough number of good enough qubits. And you’re basically saying that if I wanted a 200 qubit machine using your technology, I would need roughly 200 resonators, right? 200 cavities.

Julien: Yeah, that’s the purpose of the technology. So I think something that is interesting about that technology is you revert also to our paper in 2024. So we are able to do quantum error correction on single cavities. If you look at the performance of that system in 2024, it’s not where we need it to be in terms of real-world applications. So we still need to improve on quantum error correction and so on. The thing that is great about that system is we have a way to do this without increasing the footprint of the system. So this is the multimode code approach that I’ve mentioned. So building better codes in single cavities, for example, is one way to do this, but also we’re improving across the board. So as you’re mentioning, Yuval, so this is really the purpose of this system is really what we’d like to achieve, what we call the one-to-one ratio. So one cavity is an error-corrected qubit and scale the technology with this building block.

Yuval: If we just go into the 2024 paper, was it essentially quantum memory or were you showing quantum operations as in gates?

Julien: Yeah, so this paper in 2024 was a demonstration as a quantum memory. Since then, we have done gates that we presented at different conferences. And in our blog post, for example, in our blog page, you will see also the operation on gate of these qubits.

Yuval: Now you know what the next question is. So when can we see some gates?

Julien: What can we see in this? So we already have demonstrated gates. So you can see this on our webpage. I think the picture I have behind me is also a picture of our four qubit system. So this will be in operation this year and also something that we intend to work with partner to be operating with all these gates, but also working directly with error-corrected qubits. So this is for this year, the first release of that system and more publicly, if I would say.

Yuval: The 2024 demo also used some post-selection, right? Or was it just 100% error correction?

Julien: 100% error correction, yeah. So no post-selection. We also improved on that layer. So at that time, I think there’s one metric that we track internally is the quantum error correction gain. So by which factor are we improving sort of the performance of the hardware with quantum error correction? And now we also more than doubled on that number. So initially it was just barely making it. So I think we had a factor of 1.1 at that time. Now we’re beyond two and keep improving also the quantum error correction capabilities on these codes.

Yuval: Could you explain what that means? So let’s assume that you now, someone will test the four qubit system, maybe have some gates in it. You have built-in error correction. So what’s the gate fidelity? Is it a hundred percent or is it? Three nines, six nines, nine nines. I mean, what are we looking at?

Julien: Yeah, yeah, so we’re looking at, so depending on exactly what we’re looking into, so I think we’re talking around three nines is what we are working with right now in terms of operation on a single cavity and the two qubit, yeah, this is still improving. And also like we are adding the quantum error correction layer on top of all of this. What we think is interesting about this is the fact that we’re running quantum error correction today with these cavities. What it means is also we are improving this quantum error correction layer, but it comes with no surprise after this. I think this is how the system runs. You don’t have these things like post-selection. We’re not compromising on certain gates to allow for certain codes. So it’s the same codes that will be running on these systems. The other piece is also that we don’t have this classical layer that is usually there in quantum error correction to sort of do the syndrome measurement. So all the error correction that we’re running here is done autonomously, which is another differentiator of the quantum operation on that system.

Yuval: Three nines is nice, but certainly not enough. So does this approach take you to, what, a tera quop? How many nines does it take you to, or what do you have to do to get to the finish line?

Julien: Yeah, so I think it’s, so what we have to do is more of the same. So what I’ve mentioned earlier is, of course, the performance of the quantum hardware that we have today is not to reach usefulness. However, what we have unlocked with these cavities is also a way to improve performance with this multimode quantum error correction approach that I’ve mentioned earlier to unlock higher performance. And we have done demonstration of this. So you refer also to post-selection, which is something that we can do with the Tesseract code that enables us also to, which is sort of the two-mode version of the GKP code that I’ve mentioned earlier. And on that cavity, what we’ve shown is that we can operate with similar performance as the GKP code without post-selection, but when we turn that layer on, we’re able to shut an error channel off. And then all of a sudden, the logical performance, so the decay of the logical information, we were not able to measure it at that point, meaning that the decay was extremely low for that experiment. So what I’m saying here is that, yes, we need to improve, but we are improving by building these quantum error correction codes that are built on that artillery of multimode codes inside single cavities. So that’s the path forward for the company. Of course, we’re building bigger systems. That’s one axis of scaling. The other one is just pure quantum error correction improvement, which in each of the cavities that we’re building. And then what are we targeting? So we think we can reach 10 to the minus seven to 10 to the minus nine error rates, which in single cavity with these multiple techniques.

Yuval: When?

Julien: By when, exactly, that’s the right question. So our target for this is around 2032. So this is in terms of error rate, of course, we’re improving from now to then, unlocking these different codes with this.

Yuval: What you’re showing, what you showed in 2024, what you’re telling me now is very impressive. So I’m curious, why does it take six more years to get to what you’re describing? What’s the holdup? Is it you need just more money? Is there algorithmic issues that you have to solve? What is stopping you?

Julien: More money is always welcome, of course. But at the same time, I think this is something we should chat about also. We had a pretty capital efficient approach up to now. We believe also our technology allows for something that is more capital efficient growing from now on. I think some of the challenges that we have faced with bosonic codes also relates to the fact that it’s very different from the rest of the industry. I think a lot of the simulation tools we had to build, how we do quantum error correction for this hardware is also specific to that hardware. And this is part of the things that we have unlocked until now. And from now to the future, I think what we’re building is not only just better qubits, we’re also scaling the technology. So I’ve mentioned trying to achieve also these thousands of logical qubits that we’re aiming for these similar timeline that I’ve just mentioned is part of the journey that gets us there. So yeah, so more scaling, a lot of engineering and integration of this quantum error correction capability into scaled-up systems is what is ahead of us.

Yuval: There’s been a lot of press recently about high-rate codes. You know, Harvard released some results and others. So whereas previously one would imagine that you need a thousand physical qubits to get a good enough logical qubit, maybe the answer is we just need two, for instance. How are you thinking about these publications? Do they encourage you? Do they discourage you? What are your thoughts?

Julien: I think it, like, first of all, it’s impressive. It’s interesting. I think a couple of years back when people were looking at quantum error correction, I think it seems also impossible that we would have that discussion and reducing qubit count to these pairs. I think something that convinced us that the approach is the right one here. It’s also the trade-off between number of qubits and clock rate, we believe is good with the bosonic codes that we have right now. The other piece is also, we don’t need to rebuild the whole architecture for every new code that we’re implementing. As long as we have enough modes in our cavity, this is something that we can do on the fly as well. So no, it doesn’t scare us. I think it just shows progress. I think it shows also the importance of quantum error correction and the work that is needed to be done from an industry perspective.

Yuval: I believe you’ve transitioned from CTO to CEO, you know, a couple of years ago. What have you, how does that change your perspective? What have you done or what has the company done differently since that transition?

Julien: That’s a good, good question. So I think we expanded the leadership team also just to expand our bandwidth as a leadership team. Of course, I do more and more fundraising now, so this is just more of my work. But also, we’re building the company right now and we’re extending in that perspective. That switch of role was really just to enable us to be more scalable in that front. I think we have been successful to that point. So we’re really in a growth phase right now. So scaling the team, scaling the technology, but also just growing as an organization.

Yuval: Beyond fundraising, is there a commercial movement that you’re doing right now? Are you selling access? Are you selling computers? When you have that four qubit machine, how do you expect people to use it?

Julien: Yeah. So right now we are in discussions for a sale of a system. So that will be our first system sale. How we’re thinking about these system sales and just system access in general is we like to work with partners. So we have partners also that are developing applications on these. The machine also that we’re running with four or fewer qubits at the moment, and right now these partners have access also. Working also closely with what we call our application teams that support these external partners accessing the hardware or running any algorithms on these qubits as well. So this is how we also anticipate the next phases for commercial traction. At the same time, I think selling a ton of systems is not something that we will see from Nord. I think what you will see from us is selling more and more of these systems, but at the same time, our goal is really to have commercially viable quantum computers at the right moment. So we’re really aiming for usefulness here, and for us, this goes through unlocking quantum error correction and so on.

Yuval: You’re describing an architecture that is, well, maybe very complex to build, but potentially simple to use, right? And no need for a classical error correction there. Therefore, maybe the programming is rather simple. So what do you expect people to do with a four qubit system? I mean, obviously you can’t do something truly useful. The software sounds like it’s straightforward. What are you hoping to get from initial partners using that four qubit system?

Julien: I think, yeah, you’re right. So we won’t run any useful algorithms on those. Despite the fact that there’s some interesting angles sometimes, also there’s some specific things we can do with our present system that goes a little bit beyond this qubit world and quantum error correction. But I think the first piece is just running algorithms with quantum error correction on it, on top of it, and see how it impacts also the algorithms and so on. So it’s a good benchmark for people just to have access to an error-corrected system and us just to learn how these algorithms get to work with all the artillery in place.

Yuval: You mentioned capital efficiency. Is that by design or just it wasn’t the right time or was difficult to raise money some years ago?

Julien: No, no, no. It’s by design. So yeah, so I’ve mentioned, so we’re a spin-off of the Sherbrooke ecosystem. So there’s a quantum institute there, so it’s called Institut Quantique. So Quebec is a French-speaking nation, so everything is written in French as well. But Institut Quantique is one of the powerhouses also in quantum computing development across Canada, especially when it comes to hardware. So we’re really a spin-off of that environment. That environment also has a lot of history when it comes to microelectronics. So it is part of what we call the Northeastern Semiconductor Manufacturing Corridor that runs all the way down to Albany, ends up here in Sherbrooke and in Bromont, where fabs, but also with strong ties with the microelectronic industry, like plants like IBM packaging plants that is located in Bromont, Teledyne also has a plant there, they build the lens that are responsible for the rover that is on Mars right now. So there’s a lot of this sort of microelectronic industry, which means also that in our ecosystem, there’s a lot of tech transfer happening in this sort of sector. Quebec and Canada have invested a lot of money. In our case, it’s more than a billion of investment that we are leveraging from different pieces of infrastructure. Think of these fabrication facilities that we are using. So we have access to fabs that are also dedicated to superconducting fabrication. And it’s really our team operating this infrastructure on the ground that runs this. But we didn’t have to pay for this. Like, this was there, we’re just renting the access. And that is also true for some of the quantum hardware that we need to access, so fridges, a lot of the benchmarking facilities. So we could rent a lot of this equipment infrastructure. So this is money that we didn’t have to spend. So it really is by design. So the ecosystem in which we are is really embedded into this like tech transfer philosophy. And there’s a lot of infrastructure that was put in place that we were able to leverage. And now of course we’re growing. So that model where we rent a lot of the infrastructure has an end, but at the same time, it ends at the right time. I mean, we’ve done our demonstration and just from a capital perspective, so we’re getting money at the right time to do the right thing. And the second piece for us is really the technology. So the fact that we have this overhead reduction in terms of achieving quantum error correction and so on, means that we have smaller systems to build and this costs less in the end.

Yuval: As we get closer to the end of our discussion today, just a few quick questions. One is if the technology that you’re describing was not available, no bosonic codes, just gone, what modality would you prefer?

Julien: That’s a good question. It’s a tricky one. I think all modalities have sort of advantages and disadvantages. So beyond the bosonic codes, which I think has a really strong interest, especially as it comes to scaling technologies beyond this, we’re still a superconducting platform. The reason we are a superconducting platform is, of course, that’s a technology that was developed for a long time, so we know about it, but also it’s fast. I think that’s also something that we benefit from on the bosonics side. So the technology is fast. We keep that operation rate functioning for our bosonic qubits. And I think this will be important in the end. So I think that that’s a strong axis for superconducting qubits and so on. I do think this more standard transmon-based approach faces some challenges, especially when it comes to like connecting multiple fridges into a network environment. So there’s a lot of technology that needs to be advanced in order for the community to get there. But just if we are able to pass that wall in some sense, the fact that the technology is fast will remain a key advantage there.

Yuval: And as you think about the market evolution over the last couple of years, what has most surprised you?

Julien: I would say that the rate of evolution of the quantum error correction field, I think it’s, it’s impressive seeing, and it was a lot of the different demonstrations, whether it’s superconducting platform, neutral atoms, and now more and more people starting to bring that technology forward, to push for it. And we’re starting to, I think we will soon start to see the benefits of this in systems that will not only be playing with quantum error correction, but really start integrating that technology and scaling forward. Because like in our perspective, this is really one of the critical bricks that is in today’s most advanced systems and we really need to solve this as an industry if we want to have a useful system one day.

Yuval: And last hypothetical, if you could have dinner with one of the quantum greats, dead or alive, who would that be?

Julien: Sorry, say that again?

Yuval: If you could have dinner with one of the quantum greats, dead or alive, who would that be?

Julien: It’s funny because I was just at dinner with Michel Devoret, so Nobel Prize winner today actually, he’s visiting the ecosystem, so I’m pretty sort of on that end here. But, but all but that would have been my pick. Also, I think this Michel is not only somebody with whom we shared strong ties, but also he’s the founder of this demonstration of bosonic codes in superconducting circuits. So.

Yuval: Wonderful. Julien, thank you so much for joining me today.

Julien: Thank you so much, Yuval.

Yuval Boger is the Chief Commercial Officer of QuEra Computing.