- How to be a front-runner in today’s quantum era and beyond,
- Highlights of exciting announcements from IonQ’s product roadmap, and
- Why it’s essential to ensure access to enterprise-grade systems that deliver commercial advantage.
This is a fascinating conversation you don’t want to miss!
Watch the full episode here:
Or listen here:
Disclaimer: This show is for information and entertainment purposes only. While we will discuss publicly traded companies on this show. The contents of this show should not be taken as investment advice.
Patrick Moorhead: Hi, this is Pat Moorhead and we are back for another Moor Insights & Strategy Podcast. And I am pleased to reintroduce, probably needs no introductions, IonQ’s CEO, Peter Chapman. Peter, great to have you back on the show. It’s been a couple years. You and I have talked occasionally and I know you talked with my all-star analyst, Paul Smith-Goodson routinely, but welcome back to the show.
Peter Chapman: Thanks, Patrick. It’s a pleasure to be here today.
Patrick Moorhead: Yeah. So much has gone on since the last time we talked and not only then was the quantum rocket ship going quickly, but it keeps accelerating the industry. But I think more importantly for the context of this conversation, IonQ. Can you talk about some of the highlights since the last time we talked? Because quite frankly, you guys are crushing it, which is really cool to see, but I don’t know if everybody is aware of it. So talk about your baby here.
Peter Chapman: So we went to the market via a SPAC, which people say is a four-letter word. And I think for many of them that turned out to be the case. And prior to the merger, we put together a both financial and technical roadmap. And I think for many companies, they put together an ambitious plan, but they weren’t able to actually follow up on what it is they said. But for IonQ, we’ve been, is your word, crushing it. We’ve been doing exactly what it is that we said we were going to do, and in fact, actually exceeding it on both technical and also financial fronts. And so the company has been rewarded this year in being for most part of the year, the top 10 stocks for the year were on the NYSE, which is amazing and just unbelievable trading volume on a daily basis. It’s quite remarkable to see that. So it’s quite healthy.
And people are excited about quantum because we had said at the very beginning at about roughly 70 qubits, 65 to 70 qubits is when you could take on the world’s largest supercomputers for certain applications. And we just announced a machine, the availability of taking orders, and we actually got our first order for that for a machine at 64 good enough qubits, what we call algorithmic qubits. So people have been waiting for quantum to deliver on its promise and we’re finally getting there. And right on schedule, just exactly the way we said we were going to do it from three years ago. So we’re very happy with the progress.
Patrick Moorhead: Yeah. Doing what you say you’re going to do is critical and actually quite amazing when it comes to quantum, because there are so many things we don’t know. It’s not like you’re walking into this market that’s been around for 30 years and you’re just coming up with a different flavor on it. The other thing that’s been fun to watch is you’re a public company, which means you have to give people details, you have to give investors details, and there are very few companies that are public in quantum that are sharing also as much information as you do. And I know I appreciate it and I’m sure the investors do as well.
One of the biggest questions I get about quantum, well, I get probably top five and I think we’re going to address all of them, but it’s about scalability. I’ll talk to a vendor and it’s like, “Yeah, but we’re focused on scalability versus football field-size equipments.” It’s vice versa, but can you talk about your strategy for scalability here? Because I do believe that you have some advantages here.
Peter Chapman: Yeah. So scalability is obviously to a commercial end. So I’m going to draw an analogy to kind of the early internet. In quantum, you could think at the end when we’re all done, we expect to be able to simulate drugs inside a human body and run drug trials virtually and no longer even need drug trials, but that’s actually a quantum computer in the future. So I’m going to draw an analogy that’s like being able to download a Netflix movie in maybe 10 seconds, but we’re in a much earlier stage of the market.
And so the internet is here, aka quantum is here, but you can’t do that yet. So the question is what can you do with a slower internet? Well, you can still maybe watch movies, although maybe every once in a while it’s glitchy. You have to wait for it to buffer. You can certainly post to message boards and you can use a browser to be able to get to it, although sometimes maybe you’re a little frustrated with the speed that it takes for a page to load. So where we’re very much focused on is with these early quantum computers, what is the business value that we can unlock even though we don’t have the final product that we’d really like to have? Which is that high speed internet connection, if you will.
And I think that’s a large part of the difference sometimes between us and some of our competitors. Sometimes our competitors, we hear they’re striving for perfection, but that if you kind of did that in the internet case, we wouldn’t have had Netflix because you would’ve had to wait until you got to today’s internet speeds. So we’re trying to figure out kind of in these early quantum computers, can we make them good enough to be able to unlock a problem for a business customer? And this AQ 64, 64 Algorithmic Qubits is where we expect that to be. It’s not the end, it’s the beginning. It’s just the first place where we think where we’ll really be able to deliver large value for customers.
Patrick Moorhead: It’s interesting. The timing, I am an ex-product guy, I used to have a real job and sometimes when I hear about targeting, I mean there are multiple strategies you can take. You can say, “I’m focused on 10 years from now and that’s when we’re going to be optimal.” In fact, I heard that in a keynote this month, okay?
Peter Chapman: Yes.
Patrick Moorhead: And I’m thinking, “If I wanted to limit the amount of resources that I invested and not have people ask me a bunch of questions about that for, I don’t know, three years or five years, who knows, maybe that’s the right thing to say.” But you’ve actually come out and you are talking about commercial advantage and scale and getting people into this right now to be able to really take advantage of that hockey stick. But let’s talk about your strategy or your approach to get there. And I’m not necessarily talking about cash invested or anything financial like that. It’s more like what is your strategy to get to commercial advantage in a timely manner?
Peter Chapman: Well, so there’s multiple aspects here. One is actually financial. Obviously we do have, by taking the company public, we were able to raise hundreds of millions of dollars. That gives us an advantage. But also being the first one hopefully to market with a quantum solution that can deliver real value means that I will have an income stream that allows me to pour back into the base business. And in fact, actually this year we’re showing that quantum is a real business and we’re generating significant revenue as well. So that’s also an advantage.
Patrick Moorhead: Wait a second. Stop there right now, A quantum company generating significant revenue.
Peter Chapman: Yeah, kind of exciting.
Patrick Moorhead: I know. I mean, listen, it’s different because we probably track 20 to 25 companies and there are some who are generating revenue but won’t say how much. There are companies that are public who don’t give any sense for it, so it’s kind of a mystery. So again, audience pay attention here, a quantum company that actually has to by SEC law every quarter talk about how much revenue and give forecast out there. Sorry to interrupt you, Peter.
Peter Chapman: No, no. And since the beginning here, we started off very small, but we have been knocking it out of the park and it goes all the way back to that original what we said we were going to do. And sure enough, we’re easily exceeding those things on a financial front. And so look, a lot of people talk about different kinds of qubits and there’s a lot of competition in the marketplace here talking about our qubits, which are ions, better than superconducting and all the rest. My guess is at the end of the day, each of these different qubit modalities will have its own time in the marketplace.
Some of them will be 10 years, some 15, some 20, maybe some even 30 years away. But in the next five years, it’s definitely going to be ion-trap technology that’s going to lead the market. And then 10 years from now we will have that revenue behind us and access to public markets. And who knows, maybe sometime down the road there’ll be a different qubit modality that we like even that might replace ion traps. But in the near term, we think that this technology is the one to win. And that gives us that advantage to be able to capture the market early and generate those revenue streams before anyone else. And that’s a huge advantage.
And IonQ is the only company that’s really talking about delivering that kind of commercial advantage in the next two years. And in addition, kind of back to your original question to be able to get to scale, about half the company is working on building better quantum computers and the other half of the company is now thinking about how to build smaller, cheaper manufacturer quantum computers. And I think we’re the only company that’s at that stage, which is also a secret kind of sauce, which is required to get to much larger quantum computers.
It doesn’t matter what kind of qubits you have, at some point you can’t fit more of them on a chip and then you’re going to have to build multiple chips and you’ll have to network them together. And so we need, everyone knows Moore’s Law. There’s one side of it, which is the technical aspect, which is doubling every 18 months, but the other side was that in every generation it gets cheaper. And so we’re really focused on the cheaper side of Moore’s Law, which is in every generation. If you look, we just announced two systems. One of the systems, the AQ 35 system, Forte Enterprise is 40% smaller than the previous generation, and the next one is 50% smaller, which is our Tempo system than the previous.
And size is, in some sense, is kind of a predictor of cost because I don’t need all that other stuff that was sitting in there. I can now put it in a smaller form factor. So we’re very much focused in sometime down the road, probably 10, 15 years, quantum is going to need to be blade servers where you’re going to plug them in just the same way that we do. And you need to get these things down to the point where they’re affordable, so you can put a hundred or a thousand into a data center, and that’s what we’re working on.
Patrick Moorhead: Yeah, that would make sense. Very similar to what we see with AI accelerators today, whether it be a GPU or an ASIC, something like that. And it’s accessible by API. It’s not doing all the compute, but it’s doing the most specialized and biggest heavy lifting for that workload. So I heard you talk about three, sounded like some really cool names, Harmony, Aria, and I hope I get this right, Fort?
Peter Chapman: Forte.
Patrick Moorhead: Forte. I was just going to say Forte, but I didn’t. But what’s the difference between those systems?
Peter Chapman: And all the names are based off of music, so that’s where these things come.
Patrick Moorhead: Oh, I didn’t know Aria was there.
Peter Chapman: Yeah. Aria.
Patrick Moorhead: Aria, gosh.
Peter Chapman: Aria.
Patrick Moorhead: I’m just butchering this left and right. Your marketing people are disliking me right now on Twitter.
Peter Chapman: Not to worry. It’s all music based. And so each one of the systems has different numbers of algorithmic qubits. You could think of algorithmic qubits being the useful qubits that there’s enough fidelity in the system to be able to use those qubits for average applications. And not to get confused, sometimes you see people talking about physical qubits, but the error rate is the thing that controls how many of the qubits that you can actually use for an average application. So you’re looking at the Harmony system started with 11 qubits. Then the next system is AQ 25, 29, 35, and then 64. And so every time you double, you add one more qubit, you double its computational power that it can do.
So at 64 algorithmic qubits, it’s a two to the 64. So everyone should type into the browser two to the 64 and see what that number is. And it’s 18 quintillion. I always want to say quantillion because we’re quantum, but it’s quintillion and that’s a massive number. And to kind of give that perspective is at Oak Ridge National Laboratory, the world’s largest supercomputer, they do about 1.2 quintillion floating point operations per second. So here you’ll be able to explore a computational space of 18 in a fraction of a second. And on the one side they’re doing floating point operations, here what you’re doing is you’re exploring probability amplitude. And so it’s a little more complicated than a single floating point add or subtract or something like that.
So this is really now finally getting to the point where we think that we can deliver massive computational power to our customer for certain problem sets. And you’ve probably heard these machines are not good for everything, strangely have a tough time adding one plus one. And at the same time might be useful for solving differential equations.
Patrick Moorhead: Yeah, we have a lot of good computers that can do that. The market’s looking for things that do things that A, we haven’t even heard of before, but also solving the known, known problems. And it’s funny, a lot of the discussion, Peter, when somebody talks about, “Hey, what can this computer do?” The response is usually an algorithm that one 10th of 1% of the population understands what that might be. But you have some real customers that are doing things like object recognition, cargo loading optimizations, real customers. Can you talk about what your customers are doing right now with them?
Peter Chapman: Yes. So quantum computers are not good for everything. They’re not going to replace classical computers, but there is a subset of applications where they really seem like they’re going to excel. One of them is exploring the natural world. This was actually the original premise behind quantum computing was that the natural world is quantum. It’s not digital, it’s not analog, it’s quantum. And to be able to simulate the natural world, you need a quantum computer.
And so, one of our first applications was with Hyundai working on battery chemistry because it turns out chemistry is one of those things that very quickly consumes classical super compute time to the point where even just the modest molecules can’t be really fully calculated using the world’s largest supercomputers. So this is certainly one area where quantum computing, and we’re working on next-generation batteries with Hyundai, but you could see chemistry is in multiple areas. Things like in the future, drug discovery, materials science, it’s all those things that have something to do with chemistry.
Patrick Moorhead: Today, those are being solved, I think you alluded to, with these exascale that are driving either flops or a combination of flops and AI. And really the AI is just used to kind of narrow in on a narrower set of the problem, but what we’re talking about here is a 10x or a 100x of this. That’s the way it strikes me.
Peter Chapman: I think it may be even better than that today. If you really look at what happens in drug discovery, they try to make their best guess as to what’s going to work. We all know it costs billions of dollars to run through at the very beginning to the end to see whether or not the drug works. And on a daily basis we hear via the internet that the following drug didn’t deliver what they had hoped to. It seems to me, and I’m certainly not an expert, but it seems like that process almost, they might as well use almost random darts, it seems, to be able to choose those things.
The hope would be if we could simulate more of the chemistry inside the quantum computer, we could make better guesses and maybe if we had a powerful-enough quantum computer to actually run the drug discovery process entirely in the quantum computer. Now, that’s still probably years and years away before you can do that. But imagine a place where you could run through a million drug trials in a quantum computer and then have it spit out the best candidate. And so that would save drug companies massive amounts of money because all those failed drug trials, but speed along cures for common ailments because all those failed drug trials is time. It’s not just money, it’s time.
And so if we can speed up that process, then maybe when people talk about potentially curing cancer using a quantum computer, maybe we can choose much better molecules and drugs to be able to find the solution to those things. And so that’s really the hope on that side. The other areas are machine learning. And so you mentioned for instance with Hyundai, we’re doing object recognition. This is for self-driving cars. So we started off with a database of images that were taken from cars and to go through and do machine categorization or another word for that is object recognition. And We have seen excellent results from that. We’re quite excited about that.
What we’ve seen so far is as we add more usable qubits, the results just get better and better. We haven’t had to improve the algorithm, I’ve just had to add more qubits to be able to make it better. And so that’s really exciting. We started off with very small systems and now going to much larger systems. And then the next thing that we are working on is using Lidar 3D point cloud data to do the exact same thing, to do object recognition in a 3D space instead of a 2D space. And so it’s early days, but it looks like quantum computing is going to be excellent at machine learning.
There’s a couple of insights that we’ve already pulled away, we hope will scale into the future is one is the number of parameters that you need to be able to feed to the machine learning model is maybe orders of magnitude less than what you’d need classically. And two is that the model that you create seems to capture the signal in the data much better than what you can do classically. And then three, which is interesting, is the number of times that you have to run through the data to create the model. Turns out that it looks like that’s an order of magnitude less.
We were creating models with one customer took 20,000 iterations with the GPU and we could do it with 26 iterations on a quantum computer. And all of that is time and money in compute. You’ve probably heard these large language models that are feeding ChatGPT. The cost to be able to run that model is astronomical.
Patrick Moorhead: They’re astronomical and it’s a 10x increase in the training and it’s a 10x, grossly simplifying, 10x for the inference. There is some discussion that a major LLM creator is losing about $20 per person per month, which probably makes sense at this point. What’s really cool is a lot of people typically think of quantum computing as you can only apply this to let’s say what a national labs does or with exo flops. And it’s really cool to think that you can apply this to machine learning and what it sounds like is even potentially an LLM making it a lot more cost-efficient. And who would’ve thought that quantum and cost-effectiveness, I mean, it’s not where you would start, but if you’re doing something a thousand times faster, yeah, there’s going to be a, or has to be a cost advantage unless there’s a 1,000x cost disadvantage.
Peter Chapman: Exactly.
Patrick Moorhead: Even I can do the simple math, Peter.
Peter Chapman: It was the 8,000x less in terms of the width of the data that we had given it, thousands of times less in terms running through the data. And then the model turned out to be more expressive. It captured more of the signal than what we did with the classical counterparts. Again, early days we’ll see whether or not it scales into other ML things. My hunch is that it will, but certainly everything we’ve done so far is quite promising.
Patrick Moorhead: No, it’s super exciting and I mean, I remember two and a half years ago we were talking more in theory than I think you had a couple people kicking the tires, but here we have very large companies making big bets with you and whether it’s Hyundai or Airbus, folks like that, it’s super exciting. So we don’t do rumors on our show, we’re not the gotcha tech show here. But I would like you to share as much as you can about what we’re going to see in the future. And I know you laid the groundwork for your strategy, so somebody might put two and two together and come up with it. But I’d love to hear though in your words, what does the next generation look like?
Peter Chapman: I guess a couple of things. One is obviously with Tempo, which is the name of our system for this AQ 64, that is the system that we think that we’ll be able to really show to our commercial customer the value of quantum computing. So that will be a rack-mounted system. You don’t need to build a special building. If you have existing classical computers, it can sit right next to it. It will fit within roughly three standard racks. So compared to today’s systems, it’s quite small, fits within your standard AC and power and all the rest of that.
That system will be designed, currently is designed and under construction that it can be field supported. So one of the things that we need to be able to do is you have that system sitting in Europe. You need to be able to have a tech support person show up and the discombobulator is broken and be able to slide that out and replace it and put a new one in. So that’s certainly one aspect of it. The next one would be that you’ll start to see us network these quantum computers together. We’re actively working on networking initially within a data center so that you can, one of the great things about Quantum is usually we do networking for communication and you can do that with Quantum and it’s quite exciting because you can’t hack the communication, which is cool, but we’re doing it for computation.
And once you entangle two qubits, they don’t care where the qubits are. So what you’re doing here is you’re building a network to do entanglement from one system to the next. And this networked quantum computer just looks like one large, much larger quantum computer. So we’ll see that. We just announced a deal with AFRL for two systems where they’re working on quantum networking and you’ll start to see this now also across larger distances. So initially maybe between cities and eventually hundreds of miles, maybe even thousands of miles, and to be able to do communications and computation across those systems.
And then the other thing is it’s great that we’re going to have all this hardware, it’s going to have all this wonderful compute capability, but you need software to go with it. And so we’re very much just gearing up now to start to build the applications that we’ll be able to take advantage of this hardware. And so these are commercial applications, things like predictive parts failures and logistics applications. And we talked about some of the early drug discovery work, the machine learning applications. So you’ll hear more about those things as they come to fruition.
Patrick Moorhead: So Peter, just for clarity, you’re telling me that something that used to take half a room can fit into a rack. Is that-
Peter Chapman: That’s right.
Patrick Moorhead: Just so I got you. Okay.
Peter Chapman: Yeah, that is correct.
Patrick Moorhead: I can’t wait to see it. I can’t wait to see it because I’ve never really heard anything like that before. I’ve seen a bunch of them and we’ve, I think, all seen pictures of them but never in something that could fit into a standard data center rack.
Peter Chapman: We’re standing up a factory in Seattle and this is what they’re going to crank out. Matter of fact, we are just coming up finally getting, we’ve been rehabbing a building to be able to turn it into a factory and just in the next week or so we will finally get the occupancy permit so we can get started. And that’s what they’re going to output is relatively for our industry, large numbers of quantum computers. We’re applying the Henry Ford assembly line techniques next to quantum.
Patrick Moorhead: That’s cool. So I think we’ve gotten a lot of people excited about this and I’m sure we’ve got some people who are experimenting with quantum computing already, some who are actually deriving enterprise benefit from it, but for the most part, even large corporations have sat on the sidelines. But after hearing this and people are excited, how would they get started with you? What do they sign up for?
Peter Chapman: That one’s the easiest part. So the systems are available on all three of the major cloud providers. So Google, Amazon, Microsoft. If you have an account with one of those, which I think everyone does, then it’s easy to find us. If you go to our website, you can will walk you through how do you actually connect. You can get a Jupyter Notebook and for a couple of bucks you can run your Hello World first thing on a quantum computer to get started. And then obviously there’s lots of places where you can learn about quantum in terms of how to program these machines, all the usual places. MIT has online courses, there’s many YouTube videos, all the usual. Obviously textbooks and the rest as ways to get started in how to program a quantum computer.
In the future, one of the things that I think increasingly will happen, there’ll be SDKs available where you don’t need to know anything about quantum that if you are looking to do chemistry, the only thing you need to be able to do is be able to specify a molecule in a standard XML format and send it to one of our APIs and you won’t have to know anything about quantum or quantum logic gates or any of those kinds of things. Same thing for machine learning. The average person uses an ML library but doesn’t have a clue how it actually works on the backend. And so this will be just like that as well.
Patrick Moorhead: Yeah. And then the more abstractions, the more people can use it. SaaS providers, citizen programmers using low-code and no-code. That’s exciting stuff. And what I like is that while it’s very different from accelerated computing with let’s say GPUs and ASICs on ML and DL and LLMs, the industry understands that where it starts is not where it’s going to end. And Peter, this is just a super exciting update. Congratulations on the success that you’ve had. The fact that you’re meeting your commitments that by the way you have to make in public, on the public markets and hitting those is impressive. So hopefully we won’t go another two years before getting you on The Six Five Podcast. We can do this again, but thanks for coming on.
Peter Chapman: It’s been my pleasure. And maybe the next places we’ll do it out of Seattle, out of the factory there so we can show you a little of the factory floor.
Patrick Moorhead: I’ll tell you what, I am unaware of anybody who has a quantum factory. So I’ve seen a quantum data center but not a quantum factory. And that’s pretty cool, and I have been in many factories around the world, so that’s exciting. So Peter, thanks again.
Peter Chapman: My pleasure. Thanks, Patrick.
Patrick Moorhead: You got it. This is Pat Moorhead, Moor Insights & Strategy. Peter Chapman with IonQ. Hopefully, you enjoyed the show. If you did, hit that subscribe button, tell all your friends and families about it, share it with them. So wherever you are on this planet, good morning, good afternoon, goodnight. Have a good one. Thanks for tuning in.