Podcast with Georges-Olivier Reymond, CEO of Pasqal

My guest today is Georges Reymond, CEO of Pasqal, a quantum processor company. Georges and I spoke about full-stack quantum vendors and the historical analogy from classical computing, the French quantum ecosystem and much more.

Listen to additional podcasts here

THE FULL TRANSCRIPT IS BELOW

Yuval: Hello Georges, and thanks for joining me today.

Georges: Hello Yuval. Thank you, I am very pleased to be here today.

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

Georges: Well, my name is Georges, I'm the CEO and co-founder of Pasqal. So Pasqal is a quantum processing company. We are building quantum processors out of the neutral atom technology. I think we will have time to explain that to the audience.

So basically that's the most scalable technologies for quantum computing today. It has the world record in term of number of qubits. And this is published in peer-review paper, I mean. The number of qubits and the applications on top.

So the company was incorporated now almost three years ago, and we are not only building the hardware. We are also developing the software on top of it to be also part of the interview, but of a podcast. But I mean, if we need to powerful application, we really need to tailor them to a specific hardware.

And on a more personal point, so I'm a physicist by training. I did my PhD 20 years ago, pioneering the technology of Pasqal, and 20 years ago I would never have imagined that today we would create a company out of this idea; still we are here. It's completely amazing. And after my PhD, I went into the industry in R&D taking technology out of the lab, which is what I did for 16 years, either in startup SMEs, or even a large corporate.

Yuval: There seems to be many quantum startups in France. Do you think there's a particular reason for that, that it happens to be in France?

Georges: Yeah, you're making a good point. I think, yes. We are already four hardware startup in France, on quantum computing, plus a couple on the software side. I think it just shows the French excellence in quantum, I would say. I think we have at least three Nobel prizes in the field, and yes, I think, yes, that's part of it.

Yuval: Excellent. And I know that Pasqal has merged with Qu&Co which is, I think the software company that you mentioned. So I would love to dive a little bit into that. There's a debate, I think with customers, whether it's best for them to take a full stack vendor, hardware and software and application, or to select what some might call best of breed, "I'll find the best hardware and I'll find the best software package, and I'll find someone to write the application for me," with the thought being that quantum computing is a race, and who knows at any given time who's got the best computer, the most qubits, the least amount of noise, and so on. What is your view? How do you explain to customers that it's better to take the full stack monolithic approach as opposed to bits and pieces?

Georges: Okay. Well, I think we just have to look at history. This idea of choosing the best hardware and the best software, in the classical world, I mean, it's something very new. Still a couple years ago, we were forced to have a software dedicated to a specific hardware, even for these classical processors.

So basically it took us 50, 60 years of development to reach that point. And the first classical application, they were all hardware-specific. I remember the first processor by IBM, it was designed for implementing accounting, and only that, and it was specifically designed for these tasks.

And second, in the history of classical computing we always devise the software around the processor, not the other way around.

And I think that for quantum computing starting that, "Okay, most software is that, and now let's build the hardware," for me it's a real dead end. It's asking too much of the hardware first, and second, it doesn't leverage the specificities of the hardware.

For instance, at Pasqal, we are using this neutral atom technology, and we have the ability to arrange a qubit, the geometry of a qubit, in any shape we want. We can arrange the atoms at any distance we want, and in any shape we want. And if you are able to leverage that, then you are able to implement very powerful algorithm. I mean, for instance, our qubits are representing the shape of a graph or a molecule, and for a graph-solving problem, it's highly efficient compared to that if we do not have this flexibility. And so for me, it's very, the path, building on the strength of the processor, building the application on top of that, that's the shortest path for a quantum advantage, and maybe that in 20, 30 years, we will be able to devise hardware-agnostic software. But for me, it's really too early.

Yuval: So you're saying that through the historical analogy, it makes sense to build a full stack vendor. And if you look at classical computing, you can make a case that on one hand, Apple, very successful company, is a full stack vendor-

Georges: Absolutely.

Yuval: ... but on the other hand, of course, PCs, or phones for that matter that are not Apple phones, are more distributed. Okay, very well. How long do you think before customers can enjoy quantum advantage defined as something that they cannot do classically well enough?

Georges: Well, it's a good point. I think it will happen sooner than expected. And honestly, I can foresee that in one or two years from now. Of course it'll bring a quantum advantage for any applications, but for the first ones, I think that we can say about one or two years should be the correct timeline.

And what I mean by quantum advantage, so it is not obviously related to the accuracy, to the performance of the time of computing, it can also be providing the same results, but with less energy, because these quantum processors, they are very low energy intensities. So this is also part of the quantum advantage.

And on Pasqal, we have identified already two use cases where we think we should reach this quantum advantage at an industrial level, the first one has been devised with a global electricity utility. So it's an optimization problem, and we have good evidence, that we could reach an industrial quantum advantage with 1,000 qubit; and 1,000 qubit, we are on track to reach that by 2023.

And we also have a second use cases in machine learning on graphs. And the threshold in term of qubit could be even lower. So, we'll see. And maybe we have also a third one, but it's probably too new, so I prefer not to talk about it.

Yuval: When you speak with customers, their faced with a wide choice of quantum modalities, right? They can look at super conducting qubits. They can look at photonic qubits, they can look at other trapped ions, and so on and so on. How do you communicate to them, how do you explain to them why your approach is better?

Georges: Well, first we have the devices available. This is a huge difference, at least with the photonics qubit, because as far as I know, there is no viable quantum processor using photonic qubits. And so we have the processors, they are available, and with unmatched number of qubits. So at Pasqal we already have 100-qubit machines that we will launch on the cloud in the coming months.

And even though these 100 qubits, it's not enough to reach a quantum advantage. I have to be fair on this, still it's big enough to implement a real use case at scale. So basically for the end users, it will really learn something. We're working on the real use  case, and at the end, it will also be able to force it when the technology will be able to address its challenges in a most efficient way.

And this is what I'm telling to the customers by the way, "Today, it's too early for an advantage, but in one, two years, you will learn, you will train your people, we will be able to forecast when it will come. And so, and then the technology will be ready." They will have an edge compared to their competitors. We simply waited for the technology to be ready. And this is what I'm telling to the customers. And usually, yes, they do understand that pretty well.

Yuval: And your computers are available on the cloud, is that a private cloud? I mean, is that a scale cloud, or is that just one of the big cloud vendors? How does it work today, and what do you expect to happen in the near future?

Georges: Well, that's both. So that being said, I mean, at Pasqal, we do not claim to rebuild complete infrastructure from scratch. We are just building the layers to give access to, let's say, yes, to private customers to have access to these devices. And we also need this layer of software to be able to put our devices on standard cloud provider, such as Amazon, Azure, or even Google. And so it will be both.

Yuval: How many computers do you have online at the moment? How many quantum computers?

Georges: Online? None of them for the moment. Currently what we have in term of hardware, we have one devices, which is very close to, I will say, a lab experiment, but it's already available for implementing use cases, that's the one we used for the quantum machine learning techniques. And we also use it to develop the technology. So increasing the number of qubits, increasing the quality, the fragility of the operations.

And we have two of our devices being currently assembled. And these devices are dedicated 100% for end-users for our private cloud. So one of them is almost ready, it's currently getting alive, I would say. So probably that we will launch it on the cloud in couple weeks or months. And the second one will be available by the end of the year. So at the end of year, we'll have two device it on the cloud. And when I said to devices, operating between 100 and 200 qubits, that's a world record.

Yuval: Quantum computing has become a little bit of a geopolitical issue. There's a quantum arms race, different countries, the EU is spending money, France is spending money, US, China, and so on; all these computers that you mentioned, the three computers, will they be in Europe? Will they be in France? Does it matter to you? Does it matter to your customers where the computers physically are?

Georges: That's a good question. Well, the first one, they will be in France, because we have to bear in mind that the technology, and all technologies are in their infancy. So basically that the devices are still very fragile, so it's better to have them close to the engineers for maintenance.

So the first devices, for sure, they will be in France. In a first step, they will be available in the cloud. And I think it's the good way for scaling because it's, I think, very unlikely that the customer will buy devices from scratch without testing it, without assessing the proof of value of the computer.

And for this, I think the cloud is a nice way to go. So they can buy runtime hours, and not spend their full budget on the hardware, and start their quantum journey.

At the end things may be different. I don't know the truth, probably there is room for cloud computing because it's convenient, but at the same time, I can also hear from one of my customers, from many of them, that we have sensitive data, and it's more or less out of regrets that we will send the data over the cloud. So maybe there is also room for devices on premise. We will see; the market will tell us.

Yuval: I usually ask this question of end users, but I would love to ask it of you as well. Let's assume you were master of the quantum universe for a little while and you could control anything that you'd like outside your company. What would you like other companies to work on? What would you like customers to do to get to useful quantum applications faster?

Georges: How could I answer this? Okay, so it's not a direct answer, but I will dream of neutral atom that are easy to control with light. Usually we need several lasers. There are pretty complex cutting-edge lasers.

So if we could imagine a new academic species which is so simple to be controlled with light, that would be very interesting. And otherwise, well, I think maybe that I would like to see the large corporate would be more committed to a new technology.

I can see the large appetite, and it's growing to be honest; but at the same time, I think they can do more. They can do more. And it will definitely foster the adoption of these new devices, and also speed up the rise of his advantage. Because as of now, it's hard to tell the ideal use case that will first bring a quantum advantage. So the more we try, I will say, the more likely we will find it. So, that's a messages to large corporate, just try, it's not so expensive, and the return on investment can be pretty large.

Yuval: And what do you need the government to do, if anything? Whether it's the French government or the EU, what do you need the government to do?

Georges: Well, I think they can help with public procurement, because it's a way to help the companies, but not only by just giving subsidies, it's also worth of developing the business, trying to build real devices, test them outside the company in a real life environment, and at the same time with in an environment which is not so harsh, so you have time to test and to validate the take. So I think that's probably the nice way to help the startup, emerging, and keep growing.

Yuval: And as we get closer to the end of our conversation, you mentioned that you'd like companies to be more committed, or customers to be more committed to quantum. When you meet customers that are not as committed to quantum as you would like, why are they not committed? Is it because they don't believe it's helpful? Is it because they think it's too early? Is it because they don't have the right people to take advantage of quantum? What do you see as the primary reason that companies are not yet into quantum?

Georges: Hmm. I think that there are many reasons, but I think the main one is that they believe it's too early. Maybe it's too early, or they will have time to learn. And the technologies are changing so fast, it's completely amazing. One years ago we were the only company claiming to reach 1,000 qubits by 2023. And now I think there are at least five companies claim the same.

Of course, I believe in Pasqal, I think we'll be the first one to reach these numbers, but still the others are doing good. And it's not more a magic number, I think we will reach these 1,000 qubits by 2023, we will able to bring something to the end users and things are changing fast.

Yuval: When you have large number of qubits, let's say 500 qubits, for example. And let's assume that the application uses entanglement, otherwise it's not that interesting. And that probably means that you cannot truly simulate a 500-qubit computer, or 500 qubit software on a classical computer. So how do you debug an application that uses hundreds of qubits?

Georges: Well, I think it's a new field to investigate, of course, because at some point, yes have new competition meet the tools, and you will not be able to predict the results. But by the way, it was the same with the first numerical simulation, when people started simulating, I don't know, or designing a plane on the computer, they were facing the same issues.

So I think, yes, it has to be in investigated. And just to elaborate a bit more on my answer, there are some problems or some challenges that are difficult to compute, but checking, assessing that the solution is the right one, can be very easy. Because for instance, if you are solving an optimization problem, you come up with a solution, it can be very easy just to check that it's more efficient than the previous one.

Yuval: Absolutely. So, Georges, how can people get in touch with you to learn more about your work?

Georges: Well, we have a website. We also have, of course, a LinkedIn page. And people will find our papers, our white paper and whatsoever. And if they really want to, or really have a hands-on, we have an open source framework called Pulser, which was 100% devised by Pasqal engineers. It's a programmable framework which control the qubit at the pulse level, this is why we called it Pulser, so giving you an exquisite control over the processors. It's open source, everyone can use it, everyone can improve it. It comes with an emulator, So you can assess if your algorithm, which is devised, is efficient or not. And on top of that, it comes with tutorial, and some libraries. So it's really a nice way to start digging into quantum.

Yuval: Very good. Well, thank you so much for joining me today.

Georges: Thank you very much.