Transcript | Building a new Type of Quantum Computer with Superfluid Helium

You likely heard of trapped ion and transmon qubit, but there are other approaches that show promise for creating reliable quantum computers. One such technology is trapping electrons and pools of superfluid helium. Can it provide fast gate speeds and long coherence times? What about connectivity for performing large circuits for calculations? Find out in this episode of The Post-Quantum World. I’m your host, Konstantinos Karagiannis. I lead quantum computing services at Protiviti, where we’re helping companies prepare for the benefits and threats of this exploding field. I hope you’ll join each episode as we explore the technology and business impacts of this post-quantum era.  

Our guest today, Nicholas Farina, from EeroQ, has been in quantum computing since around 2016, so that’s a decent stretch for this young industry. He started out as an angel investor who was supporting a friend’s research project and commercialization, and he ended up cofounding the company. Is that correct? 

Yes, that is absolutely correct. I had a winding road into quantum computing. I came from the software side of tech, and I have a friend who is a professor of physics at Michigan State University. I think my relationship with him and my level of trust is what allowed me to jump into the field back in 2016 without having come from a quantum computing background, and here we are five years later. I’m still just as excited as ever to be in the field.

Why don’t you tell us a little bit about EeroQ? It seems to be a company focused on making a new type of quantum processor. 

EeroQ is focused strictly on hardware. Our perspective is that building a quantum computer itself is a sufficient challenge for a startup, so we’ve decided to focus purely on the development of the chip itself. When we started the business in 2016, people looked at us a little bit crazy because there wasn’t any way at that time that folks could see to get this type of chip into the hands of end users. Now, with the development of many big players providing cloud access to quantum computers, that pathway makes a little bit more sense, and now there’s a way that we think a company can be a pure-play chip company while still being able to commercialize their technology. 

EeroQ is the only company that focuses on this pathway — electrons and helium. I’m happy to talk more about the details, but it’s pretty much exactly what it sounds like: The qubits are single electrons that are trapped in vacuum above the surface of superfluid helium about 10 nanometers above the surface, and the qubit states are created using microwaves. It’s a little bit of a blend between an ion-trap approach in terms of using a natural element as a qubit but with the type of control that comes from the superconnecting-qubit side of the world. The idea is to essentially put together the best of both worlds, where you can have the really fast gate operations that you get from superconducting circuits along with the highly connective qubits that you get with ion traps, in addition to long coherence time. That’s all we do.  

How would you compare this to, let’s say, IonQ’s approach in terms of coherence times and other performance factors? 

 I’m always very clear to say this particular qubit is earlier in its development, so some of this is not known yet. Nobody, including ourselves, has created a two-qubit gate using electrons and helium. Of course, that said, there’s a reason that we’re doing what we do, which is that our belief is, once we achieve that, the benefits will allow this approach to catch up. Again, we don’t necessarily view it as having to exceed other approaches, but certainly, we think it can be competitive. In particular, with regard to ion traps, ion traps have excellent coherence times, so we believe that an electron and helium and quantum computer can meet or potentially exceed coherence time of ion traps, particularly with using the spin state of the electron. 

There are a couple of different states of the electron that can be used as a qubit. If you use the spin state, there’s quite a bit of research that suggests that that state can get coherence times of over a second. So, that is an order of magnitude larger than you’d get in superconnecting circuits. There’s no real magic to this. Just because electrons are so light, the gate speeds can be considerably faster than an ion trap — in the gigahertz range, potentially, which is a considerable improvement. These are the ion traps —it also has the all-to-all connectivity, or, we’d like to say, highly connected, because all-to-all, it’s always a little bit unclear if you can exactly promise that, but it has the same high connectivity that you’d get with ion traps. Vis-à-vis ion traps, we expect coherence will be excellent, as in ion traps, but the benefit that you were going to get is going to be the gate speeds in particular.  

I was looking at the way it’s laid out. It does remind me a little bit of the IonQ approach and moving around the ion, so you’re moving electrons to become gateable operations, right? 

Yes, that’s exactly right. For control, we’re using microwaves. We’re borrowing a bit of the CQED — circuit quantum electrodynamic — tool kit that superconducting circuits use. I think it’s impossible to scale a quantum computer using lasers. It’s difficult, and we think it’s potentially a little bit easier to scale a quantum computer using microwaves for control. That’s one benefit that we see as compared to the laser-based approach about ion traps.  

As you said earlier, when this first started out, any company that built a machine, you had to build everything too. You had to say, “This is how you’re going to access it.” Look at IBM — they had to build their experience online. They had to get Qiskit working. Now, with Microsoft, for example, they have other companies using their hardware, and they’re almost like a traditional cloud environment — Microsoft doesn’t build servers, so it’s the same kind of idea that you can put hardware in. Do you guys already have talks happening on the back end when you’re working on something like this, or is it way too early? 

We certainly have talks. We don’t want to put anything in the cloud before we are convinced people are going to really enjoy it. We don’t have a cloud partnership directly yet, but we have very good relationships, I think, with most of the companies that provide cloud access to quantum computing, and that’s certainly what’s in the road map — to connect via one of these, or multiple, cloud platforms. 

What you say is absolutely correct. Now, if you are able to build a novel, defensible, interesting quantum computing chip, there are many pathways now. Well, I shouldn’t say many, but there are multiple pathways and vendors that will work with you to get it into the hands of customers. I think building everything is great if you’re IBM. They have considerable resources. As a startup, though, particularly one that is trying to do something as relatively ambitious as building a quantum computer, our feeling is that focus on that — on what you do best — is important. In our case, our team’s core competency isn’t building quantum computers using electrons and helium. We’re looking for the rest of the stack to come in around that. 

One last thing I’ll say is, in addition to the big cloud vendors, there are a lot of startups that have come into the picture over the last four or five years, providing various elements, whether it’s the abstraction layer or control software and hardware. So, companies like QControl and Quantum Machines, Horizon Quantum Computing — there are all these interesting companies that are really helping to do some of our work for us in terms of getting this into the hands of end users. I think as the quantum ecosystem continues to develop, you’re going to see it become easier for a pure-play chip company to succeed.

Yes, it really is a full stack. I like to bring that up in terms of job opportunities. When people think about this, they’re like, “Do I have to be able to build a machine? Do I have to be able to trap an electron and helium, or can I actually just learn how to code on some layer of the stack?” The answer is all of the above, right? You could participate in many, many different ways.

Absolutely. I’ll just add to that: There’s a good bit of handwringing going on now, some which is I think probably appropriate about the commercialization of quantum computing as a field, not in any particular company. For better or worse, I think this is probably an unstoppable trend at this point. While that means we have to tread very carefully, and to not allow unrealistic promises to your override scientific integrity, on the career side, this means that there are a lot of opportunities opening up — frankly, even for someone in business development. 

I know a number of folks that have gone into this space over the last year or two who aren’t even particularly technical. They’re just very ambitious, and they’re willing to put in the hard work to begin to understand the field, but if you’re an accountant or you’re a business development manager, certainly, some grasp of quantum computing is important. But now that we have real businesses that are being built up, there are many opportunities for folks to get into the field of quantum computing without coming to the field with a PhD in physics.  

I think that’s great. Look at the rest of the industry: You don’t need to be a PhD in computer science to contribute. There are so many levels. 

Yes, exactly. 

I do like the idea of people being able to shift too. We come across machine learning developers that take pretty well to the quantum approach once they learn a few things. With these machines, do you see it constantly being something you’d have to keep in-house and then allow a remote access to one of the big cloud providers, or do you think one day you’ll be able to get this into a nice and tidy box that can be shipped off to some data center? 

I think the latter. I think we’re going to be able to get this into a — well, nice and tidy is a relative term.

Yes, not a house size. 

Yes. Our particular chip — one of the benefits, particularly in some of the new elements that we’re bringing to it — can be quite small. So, there is the potential to have many electrons on a single, relatively small chip. It will, however, require the need for ultralow temperatures. This is something that we are never going to get away from given our particular architecture. Architectures that are room temperature or close to room temperature, sure, I think that’s an advantage, that’s helpful, but ultimately, I’m not sure that it matters all that much, because refrigerators are currently pretty efficient, albeit expensive — dilution refrigerators, I should say. I do believe they’re going to get smaller and more efficient over time, just like progress is being made in quantum computing. 

Progress is also being made in the cooling technology used for quantum computing as well as many other scientific fields. As you start to get more efficient, smaller cooling systems, even if you don’t, you could see a situation where you have a data center that has a number of cooling devices, but inside those cooling devices, you simply have an EeroQ chip that is not unlike any other type of chip that’s prepared using CMOS and lives inside an environment. It might be a little bit bigger because of the need for cooling, but ultimately, we don’t need anything that’s going to be the size of a house, or many, many acres.  

Of course, that’s also going to be a sad moment, because for those who don’t know, a dilution fridge, that’s the beautiful chandelier that you’re used to seeing when you google what a quantum computer looks like. 

I think it will still be beautiful. They may become smaller, but I suspect they will still be beautiful. Yes, I do imagine that 20 years from now, the huge, beautiful dilution refrigerators that people often mistake for quantum computers themselves will probably be a blast from the past, just like we think when we look at something like ENIAC today.

Yes, I’m kind of a nerd here. Sometimes, I feel like a buying a bunch of copper and making one and putting it behind me in my room to have people see me on Zoom, and it’ll be like, “What’s that?”

You should totally do that. The APS March meeting is an annual event for physicists, and IBM threw a party there are a couple years ago where they had a light-up dil fridge with multiple different colors that was the party centerpiece. So, it was really fun, but you should definitely do that, and maybe one day I’ll get a dil fridge chandelier or something for my house.  

Yes. People will see me, and I’ll be like, “It’s very cold where I am.”

Very cold, yes.

I’d be dead at that moment. 

Yes, it can freeze off fingers and toes pretty quickly

Do you guys give any thought already to this new idea of interconnect — that maybe it’s not about making one chip with 1,000 qubits, but more something along the lines of a few chips that can be interacting to simulate that kind of power? 

One of the nice benefits that we believe of our system is that we believe as it begins to scale, you will be able to fit many electrons on a single chip and have them be highly connected. Part of the vision is to not be as reliant on interconnects. You certainly don’t need,for example, optical interconnects for an electron-and-helium quantum computer as you would for an ion-trapped quantum computer.

That being said, this is a very early field, and who knows? It could very well be that a modular type of system — for example, once we get to four or 10 qubits — connecting several of those may be easier than scaling to a thousand on a single chip, but ultimately, we would like to get to many qubits on a single chip. That is part of our value proposition and vision, because the technology itself is a little bit earlier compared to ion traps or superconnecting circuits. We need a reason to exist. We need to bring significant advantages to the table over time. So, allowing many qubits in a small area is one of the benefits that we envision of this particular system.  

I’m assuming there’s a whole lot of research you guys are doing that isn’t really public, but is it too early to hazard any guesses about what the ratio would be of physical qubits to logical qubits — what you’d actually end up with? 

That’s a great question. It’s really too early to make any predictions for that. It really is. Look, I push our team on that sometimes. I say, “Hey, can we think about this?” It’s like one bridge too far at this point. What we’re comfortable saying is that you can get gate speeds in the gigahertz range. We’re comfortable saying that using the spin state, you can get coherence times of over one second, which we believe is pretty extraordinary. We’re comfortable saying you can get many electrons in a small area and they will be highly connected, and that you can use CMOS to produce these chips. That’s about as aggressive as we can get in terms of positioning. 

I will also note that the ratio will also depend on progress being made by folks entirely unrelated to EeroQ. There are some architectures that require very, very high overhead —silicon photonics being one of them. We don’t anticipate that this will require particularly extraordinary overhead. Again, that being said, error correction is its own field, and the ratiothat EeroQ may have will depend not only on the quality of our own chip but also on advances being made by researchers in the general error-correction community. We’re pretty bullish on that community. There’s a lot of very, very smart folks that are working on error correction in a way that would impact the entire ecosystem of quantum computing. I do believe there are multiple variables there, but we’re pretty optimistic that the broader field of error correction will lift all boats of companies working to build useful chips. 

Yes, because the estimates have been pretty bleak at times. Just a couple of years ago, there was a paper published that said something like 20 million qubits would beneeded to crack RSA, and that was just to yield, what, 4,000 logical ones in the end. That’s an abysmal ratio. 

It is. I don’t expect that that will remain static. I will admit that it drives me a little bit nuts when people say things like, “You need one million qubits for this, or two million for that.” It’s like, “Look, we just don’t know.” Number one, you don’t what the ratio of any particular architecture is going to look like both present day and the future. Number two, you don’t know how the field of error correction outside of any particular architecture is going to progress. Number three, the state of the art of today’s quantum algorithms will be very different than that in five years. 

Again, I don’t want to make any predictions about when we might see quantum advantage, but I do think that the work on creating more efficient algorithms that can make use of quantum computers — I think it’s safe to say that that is moving ahead at a steady clip. We may be surprised about how many physical qubits are needed to do useful problems. It’s not going to be two, but I don’t think you can make a statement like, “You need one million physical qubits to do something useful with quantum computing.” I don’t think you can say something like that. 

Yes, it just makes great titles for papers — really terrifying numbers, and things like that.

It does. It’s true that you need a lot more than we have today. Even in the most perfect system, you’re probably going to need more than two qubits or four qubits. It’s definitely more than we need today, but I just don’t think you can say with any type of precision, particularly because, as you’ve already mentioned, the ratio of physical to logicalqubits is really important. I think if you say we need so-and-so logical qubits, that might be a little bit closer to an accuracy.  

Yes, I’d like to push for that. I’d like to push for a new metric that actually cites logical qubits and some kind of metric about their stability. It would be easier to understand. 

There’s a lot of desire for new measurements. In a quantum volume, IBM’s metric has caught on. Some folks have proposed tweaks to it, and certainly, IonQ has embraced it wholeheartedly, because their particular device now looks fantastic with it. 

Yes, four million. 

Yes, exactly, and where everyone else is stuck in the 200s, or something like that. I’m not sure that’s an accurate representation of where the field is today, and I’ve said that publicly — this is nothing I wouldn’t say on Twitter, but yes, I think we’ll get there. Ideally,someone who is independent will come up with a metric that illustrates the relative power of different systems. I assert obviously, fidelities are important. I think gate speed is important. Connectivity is important. There are a lot of different factors that go into the question of “OK, exactly what can you do with a given chip?” 

It will be like the old days, when we were actually concerned with gigahertz of a chip design and all that kind of thing.  

Yes, exactly. 

Building a gaming rig or whatever and IonQ, of course, they’re starting to move toward this algorithmic qubit, which is basically their own equation that tries to take all this into account. It gives you a small number — 20-something, 30-something, and that could mean a pretty large system. If everyone’s keeps making their own, I guess it will just lead to more confusion.  

I think that will happen. We’ll probably create our own too, because I think everyone’s machine is different, and I think we’ll do that not because we’re trying to mislead anyone, but because we want to provide an accurate depiction of what we think makes our particular machine useful. Now, I do hope that in the next two or three years, someone comes up with something that is great and addresses the needs of everyone. Again, that’s probably not very likely to happen, but I’m hopeful. You never know.  

You could call yours Helispeed to take the helium part into account. 

If we use that, I’ll have to give you royalties, but yes, exactly. 

Yes, I can prove I came up with it right here

Yes, exactly. It’s on a recording. 

I guess you have some ideas about funding in general, because that’s how you got started. How do you feel about how it’s shaping up? Is it turning into Silicon Valley all over again? Is there something different going on in the quantum industry? 

This is a great question. Let’s see. A couple of things — I believe that quantum computing is worth funding aggressively. I’m not sure how aggressively it should be funded,particularly on the software side now, and the hardware to some extent as well. I think if we back up one step and illustrate what is my thesis for the development of the industry, then I can back into the financing angle from there. 

I think that if you ballpark 2030 as when we’ll have a number of problems where there is quantum advantage — and I should mention that quantum computers are likely to only be good at very specific tasks; these are very specific machines — once we have found quantum advantage, approvable speed-up in some particular problem, that does not mean that all of a sudden, we’ll be able to do every type of problem that a classical supercomputer isn’t capable of on a quantum computer. It’s going to require a lot of work, and it’s going to be very specific, but I think by 2030, we will have multiple problems for which quantum computing is really useful. I also think that there will be significant economic value in those particular problems by then. Whether they be in drug discovery, material science, finance, it’s probably a little bit tougher to find true speed-up, but we’ll see. We don’t know. 

Therefore, I think by, let’s say, 2025, you’re going to start seeing a relatively brisk rate of M&A activity as a result of being in advance of that broad usefulness, but not too far in advance. I think the last five years have been very difficult for taking on too much capital just because if you’re taking on a ton of money in 2016 or 2017, then you have relatively quite a long time to get to an exit. 

Now, IonQ has proven this a little bit wrong because while not a traditional IPO, a SPAC is certainly a liquidity event, and they started raising money in 2015. By the year 2021, they had a liquidity event of a notable size. I have to get credit where it’s due. They have really pulled off a significant financial achievement. I think that’s impressive, but broadly speaking, I don’t see a ton of other quantum M&A over the next couple of years, just because we’re so far in advance. 

Now, I may well be wrong, and that may well be good news for us. If anything, I’m biased here to be more optimistic than I am, but I’m a little bit conservative in terms of thinking about when we’re going to see acquisitions really start. I’m less concerned that there’s going to be this quantum winter than I was a couple years ago, because it seems like, in terms of corporate interest and developments in the space, we’ve seemed to have hit escape velocity here, where we may have a few setbacks and disappointments, but I have a hard time seeing funding for the space just drying up. I do think now is a good time to start getting into venture rounds forquantum computing without selling your entire company. 

It was harder to say this before 2021, but now, I think we’re getting there — on the hardware side, at least. On the software side, it’s a little bit trickier, because we are quite a way away from having many use cases — even several use cases — for quantum computing hardware. To be working on the software side of the field, it is probably going to be a while until you can generate revenue in a way that isn’t based on the consulting. There are some quantum software companies now that are doing well. I’m not sure that I see a much bigger market for that over the next couple of years. That’s how I view the market and financing at this point in time.  

Yes, the consulting space in interesting right now. That’s obviously what we do at Protiviti, and we’re helping companies in multiple ways there. It’s getting them ready to explore the space, figure out what use cases might work for them, help them decide how many people they want to ramp up internally on working on these problems, because we are still just a little away from that advantage, that true advantage. People always ask me, “What if you find advantage?” I say, “Well, if I found it, you would’ve heard about it. It would’ve been a major headline.”  

Exactly. Someone asked me a couple days ago, “Have you found any use cases for EeroQ yet?” I’m like, “If I had found use cases for EeroQ, we wouldn’t be a startup anymore.” 

Yes, that’s a major moment. 

I’ll let you know. So, the length scale of it — but yes, you do bring up a good point,which is that if your quantum software startup is doing consulting, then that may well just be the domain of established consulting firms. Again, to be optimistic, large consulting firms may well acquire specialized consultancies focused on quantum computing. It’s not to say it’s a dead end, but I certainly see a large market for consulting in advance of folks actually using the technology, because there is a very steep learning curve. We do not have anywhere near a sufficiently prepared workforce to tackle these problems at the enterprise level. Even if we are 10 years away from any type of broad-scale impact, it’s going to take a long time to get folks filled up on this. I think it’s a good time to start having these conversations, well in advance of usefulness.  

I think I predict advantage by ’23, 2023

Hey, I would love to see that. There are some people in the field who would crucify me if I said that publicly. I can’t say it myself. I don’t know. I just don’t know, but I’m happy to hear you say it because, again, that would be great for the entire field, and that would be a sea change in how folks look at the technology. Yes, let’s just keep working on it, and that would be fantastic. Look, I think it’s possible. I’m happy to say that. I think that’s possible.

Yes, I’m basing it on the road maps — IBM, IonQ, their road maps. They’re both targeting about 1,000-qubit power of machine by 2023. I feel like at that point, it will be impossible to be able to simulate that with a simulator, so once you reach that point where you can’t simulate it anymore, something’s got to be found that just dazzles us. There has to be some application of an algorithm in some specific way that’s going to make us go, “Hey, this isn’t just proof of concept, this is actually going to change the world,” unlike our attempts at quantum advantage where it’s just doing a meaningless calculation of some type that we don’t really do anything with. 

Yes, exactly. Again, there are a lot of folks working on this from many, many different angles of the problem. I always like to stress that it’s not just about the hardware. It’s also about, as I mentioned, the folks doing broader work and error correction. It’s about folks working on building better algorithms. There are many, many, many people that are attacking this problem. That’s all part of getting to a demonstration of advantage. 

With that, I want to thank you for potentially one day giving us another back-end target on all the interfaces we’re going to be using. We’re going to write some code, and we’re going to point to the EeroQ Helispeed 50, or I don’t know what it will be called.

Yes, I don’t know where, exactly, the whole race will shake out, but I am very confident that we will provide something interesting to the world and get something in the market that helps solve some problems over time. So, yes, quantum computing has been the adventure of a lifetime, and it really is just getting started. Thank you so much for having me. I really appreciate it. 

Yes and if you guys want to check out the site, it’s EeroQ.com, and you can see some diagrams that explain how this technology is going to work. It’s pretty cool to check it out and stay informed here. Nicholas, thanks again. 

Thank you again. I appreciate it. 

That does it for this episode. Thanks to Nick Farina for joining today to discuss EeroQ and their new type of quantum computer. Thank you for listening. If you enjoyed the show, please subscribe to Protiviti’s The Post-Quantum World, and leave a review to help others find it. Be sure to follow me on Twitter and Instagram @Konstanthacker. You’ll find links there to what we’re doing in quantum computing services at Protiviti. You can also find information on our quantum services at www.protiviti.com, or follow Protiviti Tech on Twitter and LinkedIn. Until next time, be kind, and stay quantum curious.