← All episodes
Jean-Francois Bobier

Episode · Jan 12, 2023

Jean-Francois Bobier, BCG

Why enterprise quantum bets need to start before the hardware is ready

Summary

Jean-Francois Bobier, Partner and Director at the Boston Consulting Group, is interviewed by Yuval Boger. Jean-Francois and Yuval talk about a wide range of subjects, including HPC integration, venture capital investment advice, enterprise projects in quantum computing and quantum security, and much more.

Key takeaways

  • Enterprises that wait for fault-tolerant hardware before building internal quantum expertise risk falling behind competitors who start experimenting now with noisy devices and hybrid classical-quantum workflows.
  • Migrating to post-quantum cryptography is a multi-year infrastructure project, not something that can wait until a cryptographically relevant quantum computer exists, so cryptographic asset inventories need to start now.
  • Integrating quantum processors into existing HPC infrastructure gives enterprises a practical way to test algorithms and workflows today rather than betting everything on future hardware maturity.
  • Venture investors should judge quantum startups on their near-term path to commercial partnerships and hybrid use cases, not only on long-term promises tied to fault tolerance.

Read the full transcript on the Quantum Computing Report site

Transcript

Yuval Boger: Hello, Jean-François, and thank you for joining me today.

Jean-François Bobier: Hello, Yuval. Thank you for having me. I love your podcast, so I'm happy to be part of it.

Yuval: Thank you very much. So who are you and what do you do?

Jean-François: I'm Jean-François Bobier. I'm a partner and director at BCG, which is a strategy and management consulting company.

We created a deep tech research center four years ago, and I took charge of quantum computing with my colleague Matt Langione. I've been at BCG now for 20 years.

Yuval: And you cover a lot of topics. I think you cover the metaverse as well. But in quantum, you cover both security and computing. Is that correct?

Jean-François: Yeah, correct. Those are the two, let's say, big opportunities that we've been working on. I would love to investigate a little bit more what we could do with sensing, which I think is actually a shorter-term opportunity. So we are preparing a publication on that topic for next year.

Yuval: Perfect. So what have you learned over the last six months, whether in computing or in security? What do you know today that you didn't know in the middle of the year, perhaps?

Jean-François: So last week was Q2B, which is the largest conference on quantum computing. I had the chance to meet you in person there.

I came out of the conference very energized about the prospects of quantum computing. What I learned is that there are so many companies going after the hard problems of quantum computing. I could actually see roadmaps from Google, Amazon, IonQ, Quantinuum, and many others, and get some more confidence in the fact that we do have a shot at quantum advantage in '25, '26, which I think would be great news and a great acceleration for the quantum computing industry.

So last week was a good confirmation for me that we are on the right track and we are making the progress that we need. And there are actual technical milestones behind it that these companies have built and are executing on, and that gives us a lot more confidence, I think.

Yuval: If I were CTO of a large company, I'm getting bombarded from all directions about things that I should be doing. Quantum is one of them. And in quantum, if I had to choose one, whether I invest in quantum computing or I invest in quantum security, what would you advise me to do?

Jean-François: I must say we have a French president, Mr. Macron, who likes to say you should never make a choice; actually, you should do both. And I would probably say you should do both, especially if you're a large company, meaning that you have the means and the capabilities to take on more than one topic. So quantum communications, and more importantly, post-quantum cryptography, is going to be an important topic.

It will be mandated by regulations and standards bodies like NIST, probably as early as 2032 or around that time, with probably a due date for the final migration around 2035. And that doesn't give a lot of runway for large companies, actually. We estimate that, like it took them for Y2K, it will probably take them a decade to transition from the established standards to the new PQC standards.

Just because there are so many things in there, and so many things that have been hard-coded for so long. We have to keep in mind that RSA and Diffie-Hellman have been around for four decades, so they are very entrenched in the landscape, in certificates and so on. So it will take a long while.

So I would start, at least with the new projects and applications, making them compliant with PQC and crypto agility, meaning that I would also prepare for a future where maybe the new protocols are not sound, and if I need to change them, I can change them easily.

So that would be one thing. And obviously, you're part of the industry, so you will probably know that answer very well. I would say that if you're in an impacted industry, for example, financial services, pharmaceuticals, materials, I would definitely advise preparing for quantum advantage, which we believe will happen around '25, '26. That means a lot of people will suddenly be interested, and there will be competition for talent and scarce resources.

So it's better to get prepared now, I think.

Yuval: And by the way, speaking of these large companies, is it the same person that handles both the computing and the security side, or are there two different executives?

Jean-François: Actually, it's even worse than that. There are many executives involved, and that's probably the difficulty of quantum computing programs and roadmaps.

In large companies, for example, if you are in a bank and the benefit of quantum computing will lie somewhere in option pricing, right? You will need the people on the business side of option pricing, so the traders, the people involved in, I don't know, fixed income or derivatives, to be strongly involved.

If it's just the IT department, it won't be enough. If it's just some research or R&D team, it won't be enough. So it has to be a joint effort between multiple departments. That's the beauty of a technology that is transversal, but it will also require collaboration from everyone involved.

Obviously the main buyer will probably be the IT department, especially in banks where they have large IT budgets in the billions of dollars. So they can afford quantum computing, but it will require collaboration from others. Cybersecurity is in the purview of IT most of the time, I would say.

So that would remain in the same department.

Yuval: You mentioned that the cybersecurity transformation for quantum would take at least a decade.

Jean-François: Quite.

Yuval: How long would it take for the computing side? So I'm excited about quantum. I invite BCG to help me figure out a strategy. I pull together a team. How long before I know what I'm really doing?

Jean-François: So yeah, that's a very important question, and I think there will be at least two or three phases. First of all, obviously, there will be work around the use cases: what they are, what is feasible. That you can actually do now, and you can get a sense of which use cases are relevant for you and could come first.

Then we need to work on those use cases with your data, in your specific context, and that could take a lot of time. The nice thing about quantum computing is that many algorithms are public, actually published, which is nice. So you don't start from nothing. And there are also many open source initiatives.

By the way, I support Unitary Fund. BCG has been funding Unitary Fund, and I believe open source is really an important part of the solution. But anyway, you will build on the shoulders of giants. You will take the algorithms devised by others, and you will build on these algorithms for your specific context and your specific problems, and that will take a lot of time and iterations to get right with your data, processes, and everything.

Once that's right, and it could take years actually to get to a good level of satisfaction and usefulness, then you will need to put it in production, so you will need to integrate it within your IT, right? For example, there is a use case in automotive around getting the perfect sequence of cars in a factory.

It's an NP-hard combinatorial problem. There are systems that already do that today, so you will need to unplug them where they do this optimization, plug the quantum algorithm in instead, then plug it back into the production system and do it at the right speed. So for example, if it's a continuous process, the quantum computer will need to give answers continuously.

So that will be more of a production challenge. That's why in total it will take many years, we believe.

Yuval: One viewpoint that I heard at Q2B is that there's no such thing as a quantum software product company, at least today. And the reason is that no one is truly selling the product.

They're selling their experts to help you integrate, to help you develop the algorithm. And the same is true about hardware. If I make a quantum computer, you probably cannot use it as is. You need a lot of help in putting that together. Would you agree with that viewpoint?

Jean-François: Yes and no. To inform how I can predict how quantum computing will evolve, I always try to go back in time and make a comparison with the early computing efforts in the '50s and '60s.

And one thing that you could already note at that time is that the companies were actually providing systems. Systems that would combine software and hardware, because at the time, the chips were very limited. The RAM was very limited. The storage was very limited, not dissimilar to the situation we are in today, and everything was optimized at the bit level.

For example, if you take the Apollo 11 computer, I think not a single bit was unused. They used everything they could get out of the computer. So if you take that analogy and apply it to quantum, my prediction, and I might be wrong, is that the first systems that deliver value to customers will be integrated software-plus-hardware systems. The hardware companies that try to combine that and get the most out of their hardware, for example, in the form of accelerated libraries or dedicated software that does certain tasks, I think they will deliver early value to their customers.

And I don't believe that in the short term a clean and nice separation between hardware and software will work out. We'll just have to admit that we have too many limitations with the hardware. Long term, obviously, we'll get to the same situation we have with classical computing, where we have a clear stack and, I mean, I use an Intel processor now, but I can use an AMD processor tomorrow.

The same software will run, no problem, right? But we are not at that point yet. So I would say that, yeah, I believe a system approach will work better, and that's why we have companies like Quantinuum, who basically merged. I mean, they are a full-stack company. Or IBM, they have their own software capabilities. I could give a lot of names like that.

Yuval: Do you see quantum computers being part of the high performance computing data center, or do you see them still separate on the cloud, with whichever provider, Braket, Azure, and so on?

Jean-François: So there are a lot of discussions. There is a first discussion around technical feasibility, and then there's a question about acceptability from a regulatory and business standpoint.

So let's look at the two things here. The first is: is it feasible to put quantum computing on the cloud? The answer is obviously yes, because, as you know, quantum computing is slow-input, slow-output computing, right? You give very limited data to the quantum computer.

It operates in a large space, two to the power of N, but the inputs you give to the machine are actually not that much. It's not like Netflix. It's not 4K video or whatever. And it's definitely not something that requires low latency.

However, let me give a little caveat on that one. As you know, there are a lot of algorithms, the variational ones, that do quantum computing, then quickly go to classical and then back to quantum, and they basically iterate. That's why they are called variational.

And for those, you might want to optimize the latency of the algorithm as much as you can. Even though, honestly, you could argue that the state preparation time, the readout time, and so on are still high compared to just the networking, right? But it's still an argument you might hear.

So from a purely technical standpoint, I would say that it could actually live on the cloud. I'm not actually that concerned. But now look at the companies working in HPC settings, and I know a lot of my clients. Take banking, for example, right?

They have regulations that constrain where their data centers are located. They have to comply with very specific regulations, so data cannot go outside of certain jurisdictions and stuff like that. In other words, they can't put their computation wherever they want.

So I can definitely see banks buying their own quantum computer for at least that reason, and probably for another reason, which is around security, avoiding data that could get read or whatever. I also talked to pharmaceutical companies.

Same answer. They don't want to take the risk that their patented drugs can get out somewhere. And there's, at least among some people I talk to, a preference towards having the computer in-house. So you will see that data privacy will probably play an important role.

But you could say that for very compute-intensive variational algorithms, maybe you could make a case that, even from a technical standpoint, it could make sense to have a computer on-site. So in my opinion, we will have a hybrid world. A lot of computing will be procured via the cloud, probably seventy to ninety percent, and the rest will be on-premises.

Yuval: BCG operates on multiple continents, and I know you collaborate a lot with Matt in your Boston office.

You recently published a paper discussing the differences between the US and Europe in quantum computing, in terms of private investments, in terms of government work, and so on. I'm curious about the end users. Do you see a big difference between the requirements and the adoption rate of enterprises in the US relative to Europe?

Jean-François: So yeah, thank you for mentioning this report. On the supply side, for example, we looked at the private investments, and that is public data. If you take North America, including Canada, North America captures close to 80% of the private investments in quantum computing.

That's not counting what Microsoft, Google, IBM, or Amazon are doing. That comes on top, right? If you compare that to Europe, it's more like 10 to 15%. So clearly on the supplier side, there's a clear advantage for North America, just because of what I just said. Now, if you look at the users, I would say it's a more balanced discussion.

If you look at the early adopters of quantum technologies, a lot of them are actually in Europe. Look at Airbus, Total, EDF, BMW, Volkswagen. I think there's a perception in Europe, and that's true in deep tech in general, that Europe has kind of missed the digital wave and has a chance to catch up with quantum and deep tech.

So we will see more interest in Europe out of that, let's say, angst or worry. Now, if you take what the large US leaders are doing, it's still very awesome. I mean, if you take just banking, the two leaders in quantum computing are clearly JPMorgan and Goldman Sachs, by far, and what they are doing is just excellent.

Just to give two names where I believe that what they do is world-class in quantum computing. And the proximity to the suppliers will obviously help tremendously, right? But I think it's a fairer fight on the user side than on the supplier side.

Yuval: And I do see that in Europe there are quite a few national programs.

I mean, Germany is now spending a lot of money building a quantum computer. The Netherlands opened a quantum house; the UK, Finland, and so on. I think in your report you mentioned that perhaps many of these programs are not truly coordinated. The Quantum Flagship maybe doesn't have as much synchronization between them.

Is that the case? Did I read your report correctly?

Jean-François: Hmm, yeah. We didn't write it explicitly, but it's clear that, let's say, there's room for better coordination, for sure. I think nobody will deny it. When we did the interviews, that was pretty clear. So we could do better.

That said, I still believe that some dose of fair competition between states is not a bad thing, actually. As long as it helps move the needle and lift all the boats, I would say it's not necessarily a problem.

Could we coordinate better? Probably yes. But is it necessarily a bad thing? I don't think so. If you take the US, for example, you've got some federal government funding, obviously, but you've got a lot of state funding. For example, Illinois has its own funding.

And if you look at the states overall, you will find that they also compete for funding and to get quantum computing companies into their state. So I would say it's not that different. We could definitely do better, but I don't think it's a major problem.

Yuval: We're getting close to the end of our conversation, but I did want to ask you about investors.

I know that BCG does a fair bit of work with investors. Just between us, right, what are you advising investors to invest in today? Do you think that in 2023 we will see a lot of consolidation of smaller companies into bigger ones? What's your investment thesis in quantum these days?

Jean-François: Well, yeah, that's the billion-dollar question. First of all, we have to recognize that there are many different computing approaches competing, and I don't think we can name a winner yet among superconducting, photonics, trapped ions, neutral atoms, and spin qubits.

I think it's too early to say, "Hey, technology X is going to work." So what I recommend to investors is to have some kind of balanced strategy where they don't bet everything on one technology, because chances are that it won't be the right one, right?

Just because there are five technologies, if you pick one out of five, the chance that you're right is actually not very high, right? Even though you could say there is a lot of scientific evidence pointing this way or that way, I think it's very difficult to predict, right?

So I recommend having a balanced strategy, and at least investing in two or three different technologies and promising companies in those technologies. So that's one. Second, regarding consolidation, I think it's bound to happen. I mean, look at where the IT market is today.

It's a five trillion dollar market. And look at where the quantum computing market is today. There are just too many companies to support what we believe will be a hundred-billion-dollar market per year in ten to twenty years. That's too many players to share that.

So consolidation will happen. The question is when it will happen. There have already been consolidations; as you know, Super.tech was acquired by ColdQuanta, Pasqal bought Qu&Co, and there have been similar moves, and there will be more going forward.

The question will be more when the big techs, like IBM, Amazon, Microsoft, or Google, will make their move, when they decide that some startup has a technology they believe is important. We'll start to see more consolidation at that point.

So overall, I think it's still a good moment to invest in quantum computing. But you need to have a balanced strategy, or you need to be super sure and have information that I don't have on what technology is going to work out in the end, because that is super difficult to predict.

Yuval: So Jean-François, we are here at the dinner question.

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

Jean-François: Yeah. So if you give me that, clearly, and probably 95% of people answer this, I would say Feynman, just because I admire the guy. He was the youngest group leader at Los Alamos.

He was actually in charge of computing the size of the atom bomb, and this is how he realized that it was difficult to compute these things classically. The guy was funny. I've watched his lectures, so you could have fun talking with the guy. And I also liked his approach of thinking in analogies and trying to make sense of the equations, because it's super hard, right?

Even though he would always say that people who say they understand quantum just don't. But he was good at trying to get a good understanding of what's going on and giving some explanations. And I think he would be proud to see where the industry is today.

And I also think the first relevant use case will probably be around his original vision of quantum computing, that is, simulating nature. I think it's going to be the killer app at the beginning.

Yuval: And of the European greats?

Jean-François: European greats. Oh, that's a good question.

I would probably say Niels Bohr, but that's a lot to ask, right? The guy was just… he did the foundational stuff, right?

Yuval: How can people get in touch with you to learn more about your work?

Jean-François: LinkedIn works very well. My email address is super easy to guess, but you can click on any publication to reach me.

There's a link to contact me. And obviously, go to computing conferences.

Yuval: Excellent. So Jean-François, thank you so much for joining me today.

Jean-François: Thank you so much for your questions.