Over the course of the past decade, quantum computing (or the concept thereof) has ridden hype cycles, just like AI or the internet writ large.
In the past year, however, as governments across the globe have committed north of $50 billion to the sector, and billions more have poured into quantum startups, the technology is finding itself in the midst of a particularly dramatic hype wave.
Yet even as the money pours in, quantum remains a nascent technology that hasn’t seen any use in the real world. Among other applications, it offers enormous potential to revolutionize material discovery and molecular design, and to turbo charge clean energy tech. But how close we are to actually moving from isolated chips in a lab to bringing these real-world commercial applications to life remains a bit of a mystery.
In this episode, Shayle speaks to Bob Sorensen, chief analyst for quantum computing at Hyperion Research. They dig past the VC hype to map out the current state of quantum hardware, look at the timeline for fault-tolerant computing, and evaluate where the true performance gains lie.
Shayle and Bob discuss:
- Why recent claims of “quantum advantage” are often based on artificial constructs without much real-world relevance
- The difference between physical and logical qubits, and why reducing the “noise” is a defining hurdle for the hardware industry
- How data-driven AI models and science-driven quantum architectures complement each other differently in materials discovery
- Why 85 independent hardware vendors are far too many for the current ecosystem, and how market consolidation might impact investor confidence
Resources
- Catalyst: Quantum computing could be a critical climate solution
- Catalyst: Can AI revolutionize materials discovery?
- Latitude Media: Can quantum computing help solve the load growth problem?
Credits: Hosted by Shayle Kann. Produced and edited by Max Savage Levenson. Original music and engineering by Sean Marquand. Stephen Lacey is our executive editor.
This episode of Catalyst is brought to you by ENGIE, the smarter energy supplier. ENGIE doesn’t just provide the power to run your business — they supply the energy to move it forward, with reliable, flexible solutions built for what’s next. Learn more at engieresources.com.
Catalyst is brought to you by EnergyHub. Peak season puts every grid to the test — and the utilities that pass are the ones that built flexible capacity before they needed it. EnergyHub works with more than 170 utilities to coordinate 2.5 million devices and 3.4 gigawatts of dispatchable flexibility through a single platform designed to perform when it counts most. See what that looks like at EnergyHub.com.
Catalyst is brought to you by Bloom Energy. Bloom Energy fuel cells deliver affordable, ultra-reliable onsite power for hospitals, utilities, and data centers — at speed and at scale. Learn more by visiting BloomEnergy.com.
Transcript
Shayle Kann: I’m Shayle Kann, I invest in early-stage companies at Energy Impact Partners. Welcome to Catalyst.
So quantum computing is, I think, having a moment again. Last year, something like $12 billion went into quantum startups, about six times what it was the year before, and governments across the world have now committed north of $50 billion to the space. It feels like we’re on the cusp of something, or at least many folks think that we are.
The bet behind all that capital is a familiar one: quantum machines that can do things that a classical computer just can’t—break encryption, design drugs, and, as is probably more relevant to this crowd, simulate molecules and design or discover materials from scratch. That’s chemistry under better batteries, better catalysts, fuels, and fertilizers, and all these things that we talk about here a lot.
So far there’s really, I’d say, a few results that the enthusiasts point to. One, for example, is that Google ran an algorithm on its Willow chip about 13,000 times faster than a classical supercomputer, and they called it the first verifiable quantum advantage. A number of people have made this claim, I should say. What did Google simulate? Molecules. 15 atoms, then 28, checked against the lab.
But a problem that’s built to be easy for a quantum computer and hard for a classical one isn’t the same thing as a problem that anybody actually needed solved. So far, quantum seems to keep proving that it can win the races it sets up for itself. And the money is showing up arguably well ahead of the proof.
So that’s exactly what I want to get at. How close are we to the quantum computing promised land, and what is that land exactly? My guest for this one is Bob Sorensen, who’s the chief analyst for quantum computing at Hyperion Research. That’s coming up next.
//
Bob, welcome.
Bob Sorensen: Hi, thank you. I appreciate the invitation.
Shayle Kann: All right. I want you to start by giving me a brief history of quantum computing, and I think more importantly than the history is where are we today? Like, how would you characterize the current state of that market and technology?
Bob Sorensen: Well, the brief history of quantum starts with a physicist, a Nobel Prize winner named Richard Feynman, who basically said it’s really hard to simulate quantum phenomena on a classical computer. A classical computer is the system that we’ve all used, from the $600 million HPCs to what works on our smartphones. And those the calculations required to simulate or even study quantum phenomena on those kinds of systems can be what we in the sector called intractable. Like the age of the universe to solve a problem on a classical system, where quantum offers some significant speed-up performance potential. Basically a quantum system is a sandbox where you can play as if you’re operating in the quantum realm. And so that’s really where it started in the ’80s, and it’s progressed over the last 40 years or so to the point where there’s a number of quantum hardware systems that are available, at least for kicking the tires—not exactly commercial yet. And some of the software has actually become pretty interesting, and the key to quantum is incredible performance gains on a narrow class of applications, but some of those applications are really critical to a lot of industrial sectors, commercial sectors, and governments, and just about everybody. So it’s more of an accelerator. It brings computational capability in places that really matter for a lot of people that really care.
Shayle Kann: And I want to get back to the applications, because I’m interested in that too. But before we get there, where are we in the journey of like, it has been proven, or rather, it has been achieved versus it has not been achieved?
Bob Sorensen: You know, it’s interesting because when quantum was first introduced into the world, if you will, and I would say maybe this was five or six years ago where people started to think about, “Hey, what can quantum bring to my advanced computing workload?” The big question was when. And the sector did a very good job of saying, “We’re not there yet. It’s going to take us a little time.” And in fact, it may take a decade to get to the place where this becomes widespread or I don’t I won’t say mundane, but at least something that everybody could really take advantage of. We’re at that stage where things are moving from science experiments, if you will, from kind of one-off machines that are being designed and tested in an almost laboratory environment to productization, to the idea that in the next few years, in fact, some companies have already started, you can order a quantum system to be delivered to your facility and start to really test where it is.
Now, from a performance perspective, they haven’t demonstrated what we would call, you know, dramatic performance gains over classical counterparts. It’s still in the phase where some people would call them toy problems. You reduce a problem to its fundamental essence, you see how well it works, and you say, “As quantum computing capability follows the trajectory that most people believe it’s on, we’re about three to four years away from the rolling out of systems that will be able to have performance gains that will lead any scientist or engineer or researcher to say, ‘I’d rather do it on a quantum system than a classical counterpart.'” So we’re at this we’re on the precipice, if you will, of quantum systems becoming a real quantum utility—a real utility within an overall advanced computing ecosystem.
Shayle Kann: So is it true then that no one has demonstrated yet today, you know, a quantum computer solving a problem that a classical computer either could not or a quantum computer solving a problem that a classical computer could solve but, you know, 10,000 times faster? Like nobody’s quite done that yet, or have they done that but it’s in like a lab environment?
Bob Sorensen: It’s funny because, you know, a lot of organizations right now that are—hope to aspire to be, you know, significant quantum computing vendors are coming out with benchmark claims that in some sense are interesting but somewhat confusing. And organizations will say, “We have solved an interesting quantum problem and we demonstrated this performance is better than a classical system.” But oh, by the way, this particular test has absolutely no relevance to any compute environment. It’s an artificial construct, if you will, to say, “Hey, look what quantum is capable of, now let’s go forward.”
Shayle Kann: Can you just give me more detail? Like what does it mean to create an artificial environment that’s not relevant?
Bob Sorensen: Okay, here’s the simple example. There’s a simulation that was done a while ago, it’s called boson sampling. Now, imagine you are in Japan and you go into a pachinko parlor, and you know you have the thing where you drop a ball down and it bounces all around those little pins, and then eventually it falls into a slot. Say I drop a thousand of those and I say, “Okay, simulate” and when you’re done, at the bottom of this thing, you have the distribution of where those little pachinko balls ended up. Simulating that on a classical computer—the actual simulation of how every ball interacts with every pin and every deflection and everything—is classically intractable.
You can do it on a quantum system by basically saying, “Okay, I’ve got photons, and every time they hit something, they’re either going to break left or right with some degree of predictability.” Okay, that’s an example of a classically intractable problem, simulating a pachinko machine, versus one that quantum can do, because it draws on the essence of what quantum is. It’s about potential, it’s about probabilities and such. Okay, there’s no practical application for those kinds of things. So what you end up getting are, in some cases in the early days, benchmarks that sounded interesting but were terribly misleading. There was one organization that said they were able to exhibit a speedup on one of these arbitrarily non-functional benchmarks that was $10^{28}$—10 with 28 zeros faster than what a classical computer could do. And since these guys said, and this was in their blog, they said, “Since this basically is speedup was been greater than the age of the universe, this just proves that quantum actually operates in a multi-universe environment.” And what I like to say is scientists and engineers designing aircraft or car crash testings don’t like to hear about the requirements for multiple universes to actually do compute. So it was more of a fanciful experiment than something that actually really mattered.
What we’re looking for at this point in time, though, are demonstrated performance gains where you say, “Okay, I can run a job on a classical system that takes this long, and a quantum system can take this long.” As I said earlier, we’re on the cusp of that kind of capability in the next three to four years.
Shayle Kann: But that pachinko example, are you saying that we can do that or someone has done that today on a quantum computer, or are you saying that and yet it is not a practically valuable thing to do?
Bob Sorensen: Yeah, it was done. It was done a few years ago, that particular application was tested and said, “This is what happened.” But again, even the people running the benchmark test said, “This has no practical application.” It’s really just to show the potential of quantum writ large as opposed to, “This is the first step towards solving a partial differential equation or designing an optimization problem for, you know, a large factory configuration.”
Shayle Kann: It seems to me like just intuitively, the pachinko example shouldn’t be that hard to then translate into some real-world application. You like an optimization problem of one kind or another. Route optimization, let’s say some like really complicated route route optimizations for autonomous vehicles. I don’t know, I’m making something up. But like, the pachinko example seems close enough to me to something that I’m surprised there’s a multi-year gap between the time that you can do that pachinko thing and when it becomes a practical set of applications. Like what’s the—what am I missing?
Bob Sorensen: Really, you know, the bottom line in all of this is how the real world operates. Uh, you know, if you look at my favorite example, something called the Navier-Stokes equations, if if you’re designing an aircraft or something like that, you know you use Navier-Stokes equations to figure out exactly how airflow is going to go across a wing or a whole body, or some particular thing bouncing off the, say, the nose of the plane or something. Navier-Stokes equations date back over almost 200 years. It only became relevant when machines came along to actually deal with them.
Quantum’s only been around for 30 years. There is not a great corpus of applications that the quantum hardware base can say, “Okay, here’s what we’re here’s what we can do with this.” And not only is there not a large, long history to draw on mathematically, quantum computing algorithms are not generally intuitively obvious. They’re not something that we grasp as classical human beings, if you will. We’re too large to be considered quantum-based. So it’s the algorithm development phase of this is really quite complicated and slow, and unfortunately to my mind underfunded from both government and academic environments. Peter Shor, who’s basically come up with Shor’s algorithm, one of the things along with Feynman that really started the interest in quantum because Shor’s algorithm does something interesting in terms of breaking the current widespread data protection scheme, the cryptographic data encryption schemes that everything the world runs on, he figured out a way to basically deal with that. There haven’t been many landmark benchmarks or even applications beyond that in the last few years. People are trying, there’s a lot of work.
But again, some of these toy applications, while intuitively it seems, “Hey, there’s got to be a use for this,” translating that to a real-world application without all of the simplifying assumptions you have to make to make that work, is still an uphill battle. So it’s going to take time. And there’s a lot of research going on. But right now it was funny. I wish I had come up with this metaphor myself because I love it. A lot of people think right now that the quantum computing sector, think of it as everybody’s working on building a car in their garage, and you know they’re designing the system, they’re making it—they’re making it valuable, they’re making it efficient, they’re doing all these things, but one day they’re going to have to throw up the garage door and drive down the road. They’re going to have to do something with the car. And people are saying quantum systems are the cars, quantum algorithms and quantum software are the roads. We need to figure out once you got the car, where does it take you, and the algorithmic issue right now is probably the greater hill to climb in terms of general applicability of quantum systems in the next decade.
Shayle Kann: Is that a function of like, it has been sort of sequential and a lot of the effort historically has been on the hardware and you had to solve the hardware problem first before the algorithm is even necessary, and so we’re just like entering that new phase? Or I guess another way to ask the question is like, is the hardware good enough? I’m sure it’s not perfectly optimized and it’ll get better, but is the hardware now good enough that the attention needs to shift?
Bob Sorensen: It’s—the hardware is now entering that stage, and I don’t want to throw too many acronyms around, but the stage that we’re in right now that is starting to come to an end is called noisy intermediate-scale quantum, which means that the word N at the front, the noisy, means that quantum systems are still very, very error-prone. Uh, you don’t run a quantum algorithm once and get an answer. You run a quantum algorithm a thousand times, and you get a histogram of all the potential—all the outputs you got, and you hope that somewhere in that histogram there’s one that stands high above the rest. So it’s still a statistical activity. We call them shots. You do a thousand shots, and you hope that the answer that you that is correct appears 70 or 80 percent of the time. Okay, that’s where we’re at right now, mainly because of this concept of error correction. Every time you do something on a quantum system, there is a potential to make a mistake, to introduce some kind of error, to get something that’s wrong. And so this issue of error correction in quantum systems is probably the most pernicious aspect facing the industry today. How do I build a system that has built-in error correction?
So what is happening at this point is you have this issue of physical qubits. People say, “Oh my god, we have a thousand qubits.” They’re talking about physical qubits that are error-prone. Now, with certain algorithms and certain hardware and software techniques, you can take those and make them logical qubits, error-free qubits, pretty straightforward. But there’s a penalty to that. And in the old days, say a decade ago, you needed thousands of qubits, physical qubits, to implement just one logical qubit. There has been advances made at hardware, architecture, and algorithmic techniques that are driving that ratio down. So you have less and less requirements for large numbers of physical qubits to implement a single logical qubit. So right now the holy grail, if you will, of quantum is say a million physical qubits to implement perhaps thousands of logical qubits. And that’s where you’ll start to get real science and real applications. And if you look at all the company roadmaps, that is three, four, five years down the road still. So when people talk about qubits and you hear these numbers, you have to understand that the error correction scheme, in some sense, is perhaps a more important indicator of the quality of what that system really means. You can make a lot of qubits, but if they’re all error-prone, that doesn’t get you anywhere. So what you’ll hear is we’re moving from this NISQ era, noisy intermediate-scale, to what people are calling fault-tolerant QC, FTQC, machines that can actually deal with the fact that at its most fundamental level, errors are being made on a near-constant level.
Shayle Kann: I could be wrong about this. My guess is that that penalty that you pay, if the ratio between the number of physical qubits and the number of logical qubits is too high, then you end up with a machine that may might work if you have enough logical qubits to do whatever the task is that you need to do, but it’s like an impractical machine to use for most purposes—like too expensive, basically, and too overbuilt. Is there and you can do two things about that over time. Either you could reduce the number of logical qubits that you need to do some bit of work in the first place and then accept the ratio, or you could reduce the ratio and you could figure out how to make it 10,000 physical qubits to 9,000 logical qubits over time. Are we going to—is the intent to do both of those things, one of those things, or do we not need to and we don’t have to worry about this ratio?
Bob Sorensen: The thing is about qubit counts, physical qubit counts, and logical qubit reality, the good thing is it’s proceeding—progress is proceeding on all fronts. The guys that make hardware are producing more reliable qubits, there’s different modalities that have different error rates uh, and so from the, in essence, the manufacturing end of this, there’s progress being made. So what they’re doing at the same time is they’re making physical qubits that are less error-prone and they’re increasing the number of qubits that they can produce to put into a single QPU, a quantum processing unit. Uh, so some organizations are saying, “We have infrastructure, we can scale to a million qubits with the technology we have. It’s just a matter of basically engineering it to turn it into a product.”
Now, at the same time, we see architectural improvements—the way quantum systems are designed, and that is addressing the error requirements as well. People are becoming more innovative in terms of how you do the error correction schemes. There’s different ways to look at a whole bunch of physical qubits that are making errors like crazy and solve it—solve those errors into a logical. So there’s algorithmic things going on there. And then the software is becoming more sophisticated. Algorithms are becoming more defined so the qubits you have you’re making better use of. So it’s a confluence of a number of different factors that are all pointing to this fault-tolerant quantum computing capability in the next few years that will have meaningful science and engineering capabilities.
Shayle Kann: When it comes to energy, the best solutions don’t start with a product. They start with understanding your business. Engie works with customers to understand their goals, operations, and challenges before building solutions to their energy needs, because while Engie knows energy, no one knows your business like you. And because the work doesn’t stop once the paperwork is signed, Engie helps turn energy plans into outcomes. Learn more at engieresources.com, that’s e-n-g-i-e engieresources.com, or click the link in the show notes.
Is the expectation, maybe we don’t know, but is the expectation that this will be linear progress, or is there some point where there’s some step function? Like I think of fusion—nuclear fusion as an example where, you know, the bulk of the effort of the industry forever, and still today, is to to get to energy break-even, to get to $Q > 1$, you get more energy out than you get in. And like, the expectation is that that is this mountain you have to climb. Once you climb that mountain, going from $Q$ of $1.01$ to $Q$ of $10$, which is or $8$ or whatever the number is that’s going to end up being commercial, is supposed to be a somewhat easier task. So there is this kind of fundamental breaking point in the progress that becomes like a step function. Is that how this should look, like do we get to fault tolerance and that unlocks the world, or is it just like steady incremental drumbeat of more and more fault tolerance over time?
Bob Sorensen: I mean, you know, this is, you know, the story of advanced computing in general, this idea of Moore’s law, that things always get better, that there’s always exponential improvement. And, you know, I could—you could point to a number of different quantum computing vendors who are using silicon-based technologies, trying to really borrow from all the experience of what’s going on in the classical semiconductor world to tap into those advances as well. So if you look at what’s happening in the sector, it is not a linear progression; in many cases, it is a near exponential kind of capability. So this idea of moving from 10 to 100 to 1,000 qubits is going to happen faster uh than if you were just on a linear scale. And so that’s that’s really why there’s so much enthusiasm about quantum. It’s because not that it offers incredible capability today, but its trajectory is so much more impressive, especially when you compare it to what’s happening in classical.
People don’t talk—we always talk about the end of Moore’s law. We talk about a number of things where it has become prohibitively expensive and complicated and just confusing to have additional compute capability in the classical world. Some of the most expensive HPCs right now doing science and engineering work are costing $600 to $700 million a pop, and they require perhaps $200 to $300 million of electricity to run over a five-year period. So in some sense, quantum is saying, “Hey, maybe we can take some of the heavy lifting in terms of our capabilities because the classical world, for some of the most advanced computational workloads out there, is becoming a very, very expensive game, and available to only large government end-users and a small class of commercial companies that can foot the bill for a soon-to-be perhaps billion-dollar single system.” So quantum offers potential in terms of its trajectory that far outstrips where the classical world can go in the future.
Shayle Kann: All right, so let’s talk about applications then. You said, you know, maybe we’re three to four years out from starting to see these really hit. Give me some examples, like what are the—what are the early applications of quantum—commercial applications, valuable, practical applications that you expect to see play out?
Bob Sorensen: Well, I like to think of the applications breaking down to three different kinds of classes, if you will. The first one goes back to the Feynman thing, and it’s simulating quantum phenomena at some level so you can better understand it. And if you look at how the classical world does certain things, they have to make so many simplifying assumptions to do these quantum simulations that you just can’t even really trust the results at some deep level of understanding. So quantum does things, for example, like, “I want to design an advanced material that will be so much better when it’s used as the fundamental material in an advanced battery.” Or if I’m looking at a new drug design, “I want to understand how proteins may interact so do I have a better way of designing the particular molecular structure of a protein or some other molecule that serves a purpose in drug discovery?” Um, things like computational chemistry, “How can I better understand the actual physical interactions at the atomic level?” An example in the oil and gas sector is looking at catalyst design, “Can I come up with an interesting catalyst that will allow me to make perhaps more environmentally friendly kinds of oil and gas products because I understand exactly how those interactions between catalyst and materials interact, and I can come up with better ways to come up with catalysts that are more efficient?” So things that actually mirror the physics at its quantum level gives insights at how to do better design of proteins and molecules and all sorts of things, and that has all sorts of great applications.
The second one is optimization capabilities. Optimization is a classically intractable problem. I don’t know how many people here know about the idea of the traveling salesman problem. The idea I have X cities to go to, X distances apart, and once you get above, you know, 10 or 20 cities, it just takes too long to compute. But there are so many different kinds of optimization problems out there that are classically intractable that quantum can explore much, much faster. The idea that I’m scheduling, um, you know, I’m Lufthansa, just because I flew on them last week, I have 10,000 people that I have to make sure that every plane has a pilot, a co-pilot, a purser, uh four senior flight attendants, and two junior flight attendants. And I want to make sure that that adheres to a whole bunch of rules. These people want this day off, these people are on vacation, all these things. And think about how hard it would be to optimize that schedule so everybody’s happy and yet everybody has—is on a plane where they need to be. Okay, classically intractable problem, let’s not make it—let’s turn it to quantum. But the beauty of of—of optimization problems is, remember I talked earlier about this issue of noisy intermediate stage and the histogram that you get a series of answers. Optimization doesn’t have to be perfect to be good. The traveling salesman problem would be a great example. If I can reduce how far that traveling salesman has to go by a little bit on a quantum system—it may not be the best answer, it may not be the absolute, “Oh my god, I can’t believe we got this this amazing solution,” but it’s better than what you have on classical—you’ll use it. Why will you use it? Because maybe it scales tremendously. So now, I’m FedEx, and every morning, I take every truck in the United States and I optimize the route of every truck um, so it doesn’t—one of my personal favorites, I think FedEx a few years ago did an experiment, they routed all their trucks to minimize their number of left turns uh, because right on reds are legal, left on reds aren’t, and they cut their fuel costs by 20 percent across the fleet because every truck, every day, across the country, an incremental improvement in optimization capability can scale significantly. That’s another thing a quantum brings to the table.
The third one, which is a little bit newer, a little more speculative, and will require some—some—some interesting innovation, is all the traditional computational kernels that are the stock and trade of classical computers—finite element analysis, computational fluid dynamics, linear solvers. All those things are kernels in just about every mod sim application out there. We’re starting to see organizations play with those things to see if quantum can bring some solutions to those scientific kernels as well. Uh, my—the example I use is that Rolls-Royce, who makes very good jet engines, are starting to examine what they can do for their computational fluid dynamics programs, what quantum can bring to the table. What’s interesting, they got some cool results, but what I really liked about their experiment was they said, “The results were better than traditional quantum algorithm analysis would have led us to believe.” They were pleasantly surprised that in practice, it worked better than in theory. And so this opens up in my mind this whole avenue of rules of thumb, heuristics, shortcuts in programming where the theoretical physicist or mathematician would say, “That application may not work very well,” but the programmer or the scientist or the engineer will say, “Yeah, but I got a trick.” And so that to me is one of the more interesting aspects of quantum today. It’s the clever end-user, the guy that has a quantum system, he’s got a job to solve, and he goes, “Hey, I’ve got an idea, maybe no one’s ever tried it before,” and it might just be crazy enough to work. That kind of mentality has been the underpinnings of HPC programming since its inception—clever tricks to get around some of the theoretical problems that you just can’t wish away.
Shayle Kann: I want to go back to the first category, in which includes under it the materials discovery world, which is one that I’m especially interested in, and also I think one where there has been, or at least related to this, some of the announcements of like, “Hey, we’ve we’ve demonstrated this,” have been in that category. Like Google, I mean you could tell me is this a meaningful announcement, but Google did that—made that announcement about this algorithm they ran on the Sycamore chip, which was simulating a bunch of molecules, right?
Bob Sorensen: Right.
Shayle Kann: Um, my question is, I guess in comparing this to what we’re starting to see now with AI—so right, there’s this whole group of companies that are now using LLMs and derivatives thereof, and maybe like along with automated lab testing and so on, to do super-accelerated materials discovery. There’s one that I really like called Periodic Labs, we’ve had the founder on this podcast before, they’re trying to, you know, find a high-temperature superconductor, or maybe a room-temperature superconductor if they really get lucky. And they’re doing it with classical computing, but with AI in the loop. So as you think about what the capabilities of quantum computing would be in the realm of materials discovery versus traditional materials discovery but supercharged by AI, how do you compare those two?
Bob Sorensen: Well, see, AI is data-driven. Basically you’re training a system based on known understanding, known data. Uh, you know, if you look at things like drug discovery using AI, it’s literally looking through a catalog to meet some specific criteria. It is fundamentally dependent on what is known about a particular capability or a particular material or something, and it more or less says, “Okay, this is my best guess on how this will react.” Quantum is much more science-based, in some sense. It doesn’t require a huge corpus of information to reach its conclusions. It’s based more on the physics and the science of what it means to how a material interacts. So without data, AI is basically just uh, kind of, “We have no idea what’s going on,” where in quantum, it’s more about the computation, it’s more about the simulation. So in some sense, you can argue the two of them are complementary in some sense, and that’s really where we see AI in the science and engineering world, to be quite honest. It’s more of a helper. It helps guide humans make smarter decisions, but it doesn’t declare absolute truth.
And the other issue is when it does reach a conclusion, you can’t figure out exactly how it reached that solution. It’s—it—a large language model and its training and inference phases are completely opaque. You can’t say, “Why did you give me this solution versus this one, and if I just change a variable slightly, what will be the outcome?” Non-reproducible results, non-explainability. You need a human in the loop to say, “This doesn’t make sense,” or, “Let me—give me the top 10 choices uh for a new drug, and I’ll look for that.” In quantum, it’s more about, as I said, it’s more about the mathematical underpinnings, and again there you have explainability. How did I reach the solution, and reproducibility. If I run the program enough times, does it give me the same answer, and if I tweak something, how does that affect my outcome? So as I said, it’s—it should be part of an overall advanced computing ecosystem, each technology offering up its best capabilities with the researchers fully understanding what the weaknesses are of each.
Shayle Kann: All right, final question for you. Um, my perception is that we’re maybe in a second hype cycle around quantum computing right now, or maybe it’s a third, I don’t know, you’ve been around it longer than I have. But like, I’ve at least watched, in, you know, the sort of like venture and technology circles, excitement around quantum computing which then classically turns into, “Okay, this is harder and taking longer than we thought it would go into,” and so the tide goes out a little bit. And I think it’s turned back again in the favor of quantum computing of late, and I wonder, one, do you see the same thing? Where do you think we are in the hype cycle? And two, are we in the right place in the hype cycle? Is the world calibrated correctly as to how exciting this time is for QC?
Bob Sorensen: I’m worried about the amount of investment that’s flowing into the sector. And I consulted with one company who was—they were dirt poor, uh to be quite honest, and I love them because they were cheap and they were—they knew it. And somebody wrote them a check for multiple hundreds of millions of dollars, and I talked to the CEO and said, “What’s the plan for this money?” And they said, “No idea.” Um, you know, that’s not a good sign.
My concern here is the enthusiasm. And let me just give you my stock and trade answer here. Right now there are about 85 different organizations, companies aspiring to be quantum computing hardware suppliers—85 of them. Some of them are garnering some very big VC investments. Um, the concern here is, if you go say the HPC world, or even the PC world, um or laptops or smartphones, there have never been 85 suppliers in the history of HPC, total. There have never been 85 suppliers of laptops, 85 suppliers of smartphones. There are too many organizations out there, and I can confidently say 70 of them could go belly up in the next two years, and it ultimately wouldn’t affect the overall vitality of the sector, because the smart ones will rise to the top and the ones that perhaps slipped along the way will go away. Now, what happens when 10 companies go under, or five companies have down rounds? Suddenly the investors go, “Hmm, something’s wrong here. There’s only 80 companies vying to sell quantum systems.” I say consolidation, mergers and acquisitions, and actually, to be quite honest, culling of the herd is a necessary part of the quantum journey right now. And my concern is that will be misinterpreted as the sector is entering the equivalent of the AI winter of the late 1980s. Too much money spent, too many promises, not enough demonstrated performance gains.
And so to me, the concern is managing how the sector is going to have to realign itself to be a much leaner and meaner market sector where the winners win and the losers go away without drawing too much concern that all of a sudden governments are going to say, “Well, we spent billions on government programs, what are we getting out of this?” Government funding goes away. Um, end-users are like, “Well, I’m not going to invest millions on a technology that may be gone in three years,” and VC guys, investors, say, “This is really bad news.” So my goal—one of my goals is to try to get out the message to be moderate, be thoughtful. Um, don’t look at the sector writ large as—you know, no single activity is a harbinger of the ultimate fortunes of the quantum sector. You have to be a little more thoughtful about this. And the amount of money that’s flowing into it now is a little bit scary, especially when you talk to VCs and they don’t really understand the technology and some of the timeframes. You know, they’re looking at the next bright shiny object. And so there is some concern from my perspective that there could be some potential downturn in the sector, not because the technology trajectory and the promise of quantum is in danger, but because of the perception of what happens when 85 companies become less than 85 companies.
Shayle Kann: This is hype cycles. So we’re in another one. All right, Bob, this was super informative, really appreciate your time.
Bob Sorensen: Cool, no problem. Hey, this was fun. Thank you. I enjoyed the conversation, hopefully people will actually want to listen to it.
Shayle Kann: Bob Sorensen is the chief analyst for quantum computing at Hyperion Research. The show is a production of Latitude Media. You can head over to latitudemedia.com for links to today’s topics. This episode is produced by Max Savage Levenson. Mixing and theme song by Sean Marquand. Anne Bailey edits the video version of the show. Stephen Lacey is our executive editor. All of our episodes are now on YouTube. You can subscribe to Latitude Media for episodes, and you can find the audio version of this show anywhere you get your audio podcasts.
I’m Shayle Kann, and this is Catalyst.
