Members Only Content

I enjoyed recently chatting once more with Tiernan Ray of The Technology Letter about quantum computing. We’ve checked in on quantum several times during the past half decade (see here, here, and here), as the field has generated a flood of financial media headlines and has attracted more than a few speculative investors during that time.

Yet one thing that still remains perpetually out of reach is commercialization. For all of the promise that quantum holds as a technology, it remains very limited in its commercial impact. And while large companies with deep pockets such as AstraZeneca have dedicated budgets for smaller-scale R&D experiments, quantum remains stubbornly stuck in the earliest stages of the adoption curve.

Investors remain optimistically convinced that quantum will cross the chasm reach mass-market adoption. The pure-play quantum companies who are publicly traded — including IonQ (Nasdaq: IONQ), Rigetti (Nasdaq: RGTI), D-Wave Systems (NYSE: QBTS) and others — are selling for an average of 260x their forward sales expectations. Those can safely be considered “nosebleed valuations” that put even the current AI bubble valuations to shame.

But is there a middle-ground between the hyperbolic narratives of “quantum leap” and “doomed” that more accurately describes what’s going on? What metrics or factors are the most important to the companies who are actually beta-testing quantum projects? What economics would we need to see before quantum becomes affordable for smaller companies? And what role is the government playing in this commercialization process?

Tiernan and I discussed those questions and more in our conversation.

We began with a review of the problems quantum faces (at 1:47 in the video) – Moore’s Law allowed for six decades of classical computing innovation that was building upon 0s and 1s. Quantum follows very different physical rules which generally involves a trade-off between speed and fidelity. Quantum computers can either be fast or stable, but not both.

Tiernan describes that scalability remains elusive — especially when it comes to linking linking multiple qubits together to create a quantum circuit. We are still in “the age of the transistor” when it comes quantum computing and these are hard problems to solve.

Next, we discussed “Is Moore’s Law Dead?” (11:45). Transistors are now smaller than 2 nanometers in size and there are diminishing economic returns that come from manufacturing smaller transistors for chips. Creative new approaches such as chiplets or advanced packaging are playing an increasing role at making chips more powerful. But quantum could offer a necessary and much-needed step-change for the most computationally-heavy workloads.

We went on to discuss the technology roadmap (at 16:08) and the different approaches that companies are pursuing when it comes to quantum computing technology. IonQ and Quantinuum are using trapped ion technology to manipulate individual atoms, which is showing promise in stability but is limited in computing speed. Google is pursuing superconducting technology to manipulate electron spin states, in an attempt to minimize external interference. PsiQuantum is pursuing photonics and optics, which would by far be the fastest but is much less stable. All of these approaches are still very early on the technology roadmap.

We then looked at how companies are actually using quantum today (24:02). AstraZeneca recently set up an experiment for drug discovery, using an IonQ quantum computer alongside NVIDIA’s GPUs to model the molecular configurations for drug discovery. Even though quantum played a role in accelerating the experiment’s timeframe, it was still the GPU cores that were doing most of the computing work. Perhaps unsurprisingly, Jensen Huang and NVIDIA are supportive of quantum R&D projects — which they often end up benefitting from.

We dug deeper into an economic review (28:10), comparing how much is costs to buy or to rent a quantum computer. Anyone can rent quantum computing time on Amazon Braket today for $7,000 per hour, which offers a fully-dedicated IonQ Forte system. Yet this is still more than 300x more expensive than renting dedicated GPUs from CoreWeave for $22 per hour. From a commercial perspective, quantum computing will play a role to speed up the timeline of large R&D projects. But it also appears that we’re still a decade away from it being economical for smaller businesses.

We changed gears (30:05) and then discussed is there were certain places where a quantum advantage could be particularly important. There is the potential for a quantum computing arms race, where well-funded nation-states or even individual bad actors could benefit from breaking cybersecurity encryption or for military applications. Tiernan notes that even though several companies are suggesting that RSA encryption will soon be cracked by quantum computers, we were more likely still several years away from this actually happening.

In the final segment (35:07), we reviewed several of the publicly-traded companies and discussed whether quantum computing is currently investable and the potential of future acquisitions.

Thank you Tiernan for sharing your insights with 7investing! You can see more of Tiernan’s tech coverage in his publication The Technology Letter.

Please do your own investing diligence. Hosts or guests may have positions the companies discussed. See all of 7investing’s stock recommendations free for your first week.

Transcript

Simon Erickson  00:00

Hello everyone, and welcome to this edition of our 7investing Podcast, where it’s our mission to empower you to invest in your future. You can learn more about our long term investing approach and see all of the stocks we’ve ever officially recommended at 7investing dot com slash subscribe.

 

Simon Erickson  00:17

My name is Simon Erickson. I’m the founder of 7investing. I have a very exciting guest for the program here today, one I’ve been looking forward to having back for quite some time. Tiernan Ray is the founder and the writer of the technology letter. He’s been following the tech industry for several decades. I always consider him far ahead of the curve when it comes to learning about new and innovative technology trends. Tiernan, I say this every time, but I’ll say it once again. It’s a true pleasure to have you back again for the 7investing Podcast.

 

Tiernan Ray  00:44

It’s great to be here. It’s great to be speaking with you. Thank you. Simon

 

Simon Erickson  00:49

Tiernan, we’ve chatted quite a few times about the topic we’ll be discussing today, which is quantum computing. And again, the reason we follow this is because we want to kind of go where the puck is heading. We want to look at things that are three or five years out. Seems like quantum might always perpetually be three to five years out. But I’d like to take a look at kind of where does this field stand to set the table for the conversation.

 

Simon Erickson  01:10

Let’s talk first about some of the challenges that quantum computing is currently facing. I know that the technology letter you just published a piece on this today that I wanted to dive into. And then maybe a little bit farther down, we’ll talk about the technology roadmap and some of the companies that are publicly traded, but, but first and foremost, Tiernan, you know, you and I chatted about quantum computing back in 2021 It was the summer of SPACs. Everybody was so excited. You know, there was a ton of enthusiasm about this, the hockey stick projections, and everyone was just super enthusiastic about getting into Quantum. It feels like it kind of calmed down in the years past that, but the enthusiasm is back again today. Maybe let’s first discuss about the piece that you just wrote for the technology letter about there’s still a lot of challenges and issues that quantum is facing today.

 

Tiernan Ray  01:57

There are Simon and the biggest issue is scaling. You know, in semiconductor devices, PCs and servers, you’ve had six decades of scaling, which just means that integrated circuits get more complex, they have more transistors. And as the machine gets bigger, basically you can do proportionately more work. And so you always measure progress in electronics based on can you do more work with the latest generation of the technology? And we’ve seen that for six decades with Intel Xeon processors, with AMD processors. We’ve seen that with Nvidia GPUs quite strikingly in the last few years, going through the various iterations of Vera and Rubin, each of them getting more powerful because of continued progress in semiconductor physics and that amazing machinery of Taiwan semiconductor manufacturing. So the various companies, IonQ, Rigetti, D, wave, have not shown that they can scale in the same way. They simply do not have a track record of building bigger and bigger chips, if you will, by whatever means they build chips that are quantum they simply don’t have a track record of steadily increasing the number of logical functions they can perform so that they get a proportionate payoff.

 

Simon Erickson  03:18

Okay, and I want to, I want to go one step upstream of what you just described, because last time we chatted, you recommended a book which I have on back order at my library. They have not been able to find it yet. They’re trying to send it from Dallas. But you said Revolution in Miniature was the book that kind of described the six decades of progress and innovation we’ve had in transistors and integrated circuits, and then, you know, all the semiconductors we have today, but it started at just the simple binary building blocks of zeros and ones on and off switches that we put in integrated circuits. You described last time we talked about we’re still in the transistor age of quantum, where we’ve got qubits. We can kind of show some different physics, you know, quantum physics, as opposed to, kind of the traditional binary physics for classical computers. But it seems like we’re still super early. We haven’t even figured out how even figured how to build these to connect, to connect with one another, right?

 

Tiernan Ray  04:06

Just yet. That’s right. Simon, in the 1940s the transistor was really refined to be an alternative to the vacuum tube. All computers at the time, the ENIAC and various room size computers, were built from vacuum tubes. It was several years later before we got the integrated circuit, where, instead of people assembling by hand a transistor to try to make several series of transistors, which you need if you’re going to do a logic function in a computer, instead they could fabricate them as a single planar device, the integrated circuit. So it took from, let’s say, 1947 until 1961 62 63 once they had the functional thing, it took them another 15 years or so to get to integrated circuits where you had Moore’s Law and you could then sort of predictably shrink feature sizes down sub micron eventually, right over six decades.

 

Tiernan Ray  04:59

That was the key was to build multiple devices, quote, unquote, in a single planar device that you put through one process, which is the whole revolution that we’ve had. And so we’re at the transistor phase. And the funny thing is, before they were used in integrated circuits, transistors revolutionized radio, right? You had a transistor radio, which was a breakthrough device. I don’t think we’ve even seen yet, the transistor radio equivalent of quantum, where manufacturing a couple qubits produces some thing that you say, Wow, we didn’t have this before, and now we have it, you know, in the sense of a killer app, where the transistor became a ubiquitous thing that was carried by everyone, that was suddenly portable audio. So we’re not even at the transistor radio phase of the transistor era in electronics, as far as quantum is concerned.

 

Simon Erickson  05:51

Yeah, that’s right. And something we should always remind everyone listening is that a quantum computer is not just a faster classical computer. This is completely different physical laws, where we’ve got block spheres. We’re not just managing, you know, zeros and ones, electrons on or off, going through an integrated circuit. We’ve now got things like trapping ions. We’ve got spins of electrons or even photonics. I mean, it feels like there are a bunch of different technologies competing to become the transistor or the radio of the of the of the Quantum Age.

 

Simon Erickson  06:20

From what I’ve seen, at least in the roadmap, roadmap of the companies that have disclosed that they’re doing work on quantum right now, Tiernan, and you mentioned this as well, it seems to be kind of a trade off between how fast do you want the quantum computer to be versus how stable Do you want the quantum computer to be. Can you talk a little bit about the difference between, quantum speeds and coherence and the challenges that the roadmap that these companies are technically trying to solve these problems?

 

Tiernan Ray  06:46

Yeah, you need to have increasing numbers of the transistor equivalent, the qubits, but you also need to be able to have those qubits last for an actual period of time in which you can conduct an operation before you make a measurement of the quantum operation. So all of the goodness of quantum exists in the realm where there’s a superposition of multiple qubits, and that is not accessible to us until we measure it. And in order for that reaction of the qubits to take place, you need what’s called coherence time before they decohere. They decohere because of interference from the real world, from cosmic rays, other sorts of things.

 

Tiernan Ray  07:28

And so as you build up these qubits, you’re right, fidelity matters. You’ll hear a company such as Ion Q talk about, we not only have a plan for more qubits, we’re going to be the most fidelis qubits, meaning they will have a coherence time that will allow them to run each individual qubit long enough that it can interact with other qubits, and there could actually be a logical operation. So fidelity becomes another dimension of this in order to get some kind of stable operation.

 

Simon Erickson  08:01

Okay, and so the bigger picture, reason we would want to compute a quantum computer, as opposed to a classical computer, is basically, you know, we built six decades of coding and doing these operations with binary. You know, everything was just these operations that you ran in series and then parallel with GPUs. But it was still kind of line by line of code, trying to ultimately come out with an output of some input of calculations we were looking to do.

 

Simon Erickson  08:24

Quantum is significantly faster at running multiple calculations from many inputs at the exact same time. And with the combination of coherence and superposition, as you just mentioned, that would be a completely game changer, step change in terms of calculating things much, much more powerfully and complexly than a traditional computer.

 

Simon Erickson  08:42

Do I have that right, Tiernan? Is there any other reason that a quantum stands out versus a classical?

 

Tiernan Ray  08:48

That’s the theory, Simon, you’ve articulated the theory well, because you have to increase the number of transistors based on the number of values you have to compute. If you have 1000s of values, you have to have two to the n minus one. So there’s an exponential increase. There’s actually an exponent based on the number of values. As n gets bigger, you have to have more and more transistors to have more and more logic gates. So and so you start building out more and more, either larger circuits or taking more steps of the same circuits, and so it runs longer in time. And so that’s where you start to reach a ceiling on what a normal computer can do with the superposition in theory of qubits.

 

Tiernan Ray  09:32

You could run all 1000 values in one operation on the same collection right of qubits, and so you could basically do simultaneously the work you have to do in time or in space with these traditional computers. That’s the theory.

 

Tiernan Ray  09:47

The issue, of course, is not all problems are equal, and so if you do something more efficiently, it could still take a long time. You could still not get the correct answer. Because when you measure the qubit, when you go from the quantum realm to I need a real one or zero. That’s in my traditional circuitry that you have this stage called measurement, and you have a possibility that you’ll get a probabilistic answer that’s not correct.

 

Tiernan Ray  10:16

It’s a little bit like, you know, you go to chat GPT, and it seems to know a lot, until it has what could be called confabulation or a hallucination, however you want to call it, it comes out with the wrong answer, and so all of the power suddenly is in question. And so you have a probabilistic element that’s sort of analogous in quantum which means that you might run this more efficiently, but still have to run it for a long time if it’s a big problem and still get a wrong answer.

 

Tiernan Ray  10:43

So everything you said is is correct in theory, but in practice, there’s many, many caveats to whether it can work.

 

Simon Erickson  10:53

Yeah, great points. Some of the interesting quotes that others have mentioned about this space and the challenges about accuracy, I’ve saw that written that “there were more red flags and a People’s Liberation Army parade”. A clever one there. And then the “quombies”, the quantum zombies, the undead narratives of quantum advantage.

 

Simon Erickson  11:10

Is it even going to work? I think, is the question that we’re training it to solve problems we know the answers to. But if you’re actually going to unleash it the way that it should be intended, there’s no guarantee of accuracy.

 

Tiernan Ray  11:20

Right, there isn’t. You’re still trying to find out what which problems might really receive a boost, because it’s not true that you can immediately apply the theory that you explained earlier and get a significant, meaningful improvement in efficiency and processing for every single problem. So we’re still finding out which problems will yield themselves. As you could call it the killer app, if you want.

 

Simon Erickson  11:46

I do want to talk about the killer app in just a minute. Before we get past this segment, I want to ask you, is Moore’s law dead Tiernan?

 

Simon Erickson  11:52

You know, we’ve been saying this for decades that, you know, transistors were going to reach, you know, we’ve gotten so many economies of scale, and Moore’s Law was so good for so many years. Maybe they’re slowing down its diminishing returns as we’re now sub two nanometers. I believe the sub one nanometer now for Taiwan Semi, in some instances. Are we getting past this? Or is it advanced packaging, you know, stacking chips on top of each other, and other things that we’re going to keep seeing smaller and smaller transitions. Or do you think that by necessity, we’ve got to find a new architecture?

 

Tiernan Ray  12:20

Moore’s law is dead.

 

Tiernan Ray  12:22

I think Intel would never admit that, but it’s dead in the sense.

 

Tiernan Ray  12:25

Simon that it doesn’t predictably deliver the kind of scaling that we saw for six decades, where every 18 months, maybe 24 months, but usually 18 months, you would get a doubling in the number number of transistors, which meant you would have the same computing power for half the money. That’s how your PC got cheaper. Or you would have double the computing power for the same money. That’s how you know the power users would buy the latest Intel Core series, and suddenly it’s a faster PC. So we don’t get that kind of predictable improvement every 18 months.

 

Tiernan Ray  13:00

And that means it’s not a law. It’s not a law. If it’s I mean, it was never a law of nature, but the word law was meant that it predictably improved, and if it doesn’t predictably improve, then it’s not a law. What you still do have are improvements. Taiwan Semiconductor does find new ways to do things, as do all the fabs global and tower semiconductor. But you’re right.

 

Tiernan Ray  13:22

You now start to move to, well, a decade ago, AMD pioneered chiplets in mass manufacturing, where we can’t build enough big chips, so we build smaller ones, and we assemble them on an on a substrate as if it’s one big chip. And that chiplet approach totally took off in the industry, and you’re seeing that spread. And now you’re also seeing packaging as you point out ways to stack die with one another, stack them either horizonally, either lay them out horizontally, like chiplets, or stack them vertically. Which is what is happening, for example, with DRAM memory at Micron and others. Which is, you know, if we need more storage of bits. We have to go vertical now.

 

Tiernan Ray  14:03

So, yeah, you’re absolutely right. The packaging age is upon us, and in part of it is because you can’t reliably just turn the dial and shrink the transistors year after year.

 

Simon Erickson  14:14

It’s getting harder. It’s requiring more creative and perhaps even more expensive solutions.

 

Tiernan Ray  14:19

It is. And I think, I think it’s brilliant to bring it up, because it, it is a gating factor for everyone, including Ion Q, Rigetti, D wave, IBM, Google, Amazon, Microsoft. Anyone who wants to fabricate a quantum device is having to go to Taiwan, Semi or someone else, and deal with these same scaling issues.

 

Tiernan Ray  14:42

Now, the one thing that the the IonQ’s of the world will claim is, well, we can actually build these qubits in an older process technology, because every time that the features get smaller, whenever they do a Taiwan semi to make the. Latest cutting edge Ruben GPU. Every time that happens, there’s a bunch of the old stuff left that is fully depreciated but can be used to make stuff, and usually it’s used to make stuff that’s less demanding.

 

Tiernan Ray  15:13

So the chip that’s inside your wireless earbuds might be an older technology. It might be something crazy like quarter micron, right? 250 million, you know? And so that older technology that’s fully depreciated can sometimes be a cheap way to make other devices. And there’s a claim by the quantum folks, we can use the older stuff, and so we can use the cheap, depreciated technology. I think that also is a promise that remains to be proven.

 

Tiernan Ray  15:42

But generally, broadly speaking, your point is well taken that we’re dealing everyone’s dealing Nvidia, AMD, Intel. Everyone is dealing with the issue of Moore’s Law and the issue of how to continue to get more out of the same semiconductor manufacturing technology, and that includes the quantum companies that have to, at some point, benefit in some way from the technologies for manufacturing that are out there.

 

Simon Erickson  16:10

Okay, so let’s talk a little bit more about those technologies. Well, let’s go back to the future. Here. We’re going back to the year 1957 but it’s actually the year 2025 you know, where we’re kind of developing the roadmaps of these different approaches to quantum computing, because back in 57 we were trying to create a CPU, an integrated circuit built off of binary and the transistors and all of the ways that companies, you know, built Moore’s law over the next six decades.

 

Simon Erickson  16:33

But now quantum computing is not a universal term. There are different approaches that we’re starting to see some companies put their bets on and invest in just a couple of examples. Again, this is kind of the trade off of what is possible and maybe perhaps easier and more stable, versus do you want to take a swing for the fences and do something that’s very difficult, that could be very fast, and solve problems that the other technologies could not.

 

Simon Erickson  16:58

Just to keep you up with a couple of these, Tiernan, but we’ve seen now Ion Q has gone public. It is a publicly traded company that is trying to do trapped ion technology. This is where you’re manipulating atoms in a way that you can actually monitor and measure the quantum outputs from them. We’ve got super conducting technologies, where there’s spins of electrons in a similar way that you’re manipulating. And then we’ve also got a new field of photonics from privately held companies like Psi Quantum. Psi Quantum is pursuing photonics, which could be much, much faster, but don’t have the same stability as ion trapping turning.

 

Simon Erickson  17:34

I know this is early. I know we’re beginning our first inning, but I have to ask you the question, which one of these do you think shows the most promise if you’re going to bet on one of these horses in the race? Which one are you going at? Which one are you going after?

 

Tiernan Ray  17:44

None of these companies has shown that they can scale any of these approaches. The larger companies you’re mentioning Simon such as IBM and Google. Google is probably the poster child for the superconducting approach, and they’ve been well publicized. They are at a stage of a handful of qubits, so they don’t have the equivalent of an integrated circuit. And that’s very clear that’s out there. Ion Q, in their case, is now saying, after they acquired Oxford, Ionic, a British company, this year, for about a billion dollars, that they’re going to dispense with what was an optical approach, where they use mirrors, modulators, laser light sources, and now they’re going to put it into an integrated circuit that will use radio frequency waves. So you can imagine it might be a little bit closer to something like a Qualcomm base band, where the chips function is to control radio frequency signals. And the point of that is, you can create a magnetic field that can manipulate the trapped ion, similar to how up until now, Ion Q has been using lasers. Right? You have lasers hitting the trapped ion and manipulating it. Now they’re going to do it in radio frequency. And the idea is, it’s easier to make a radio frequency chip than it is to make a bunch of all kinds of optical components.

 

Tiernan Ray  19:03

And so all of these companies have these wonderful different novel approaches, but again, none of them have shown that this is scalable. Google has not shown that superconducting is scalable in terms of getting to potentially 1000s of qubits in a single device, like an integrated circuit. But all of the promises from Ion Q about radio frequency are based on the work of a startup, Oxford Ionics, that to date, has not produced a massively parallel device that could lead to hundreds or 1000s of logical operations in sequence, which is what a programmer needs if they build an algorithm.

 

Tiernan Ray  19:38

So I hate to punt on this, but I have seen none of these companies demonstrate that any of these approaches are scalable. I see lots of things in PowerPoint that say they promise they will be scalable. The line always goes, you know, up and to the right about how they’re going to reach several 1000 qubits by certain. Year right now. 2030 is the year when everyone reaches scale, but it’s in PowerPoint. They haven’t demonstrated it for real.

 

Simon Erickson  20:07

How important is the number of qubits that these companies are reporting? Is this something that’s an important metric to you, or is this just something that’s a neat R&D milestone for them?

 

Tiernan Ray  20:16

It’s not important. It’s a way to market what they can do at the moment. And if you have 10 qubits and the other person has five, you have twice as many qubits. You know, if you have 100 qubits and the other person has 50, you have twice as many. So it’s bragging rights, but it’s while it is debatable how many qubits you need. And I think we’re some of that we’re finding out in research. It’s pretty clear, with the work we’ve seen done so far, that 10s of qubits does not yet yield a lot of interesting, broad based applications. And so it’s pretty clear these companies are below, sort of below, a floor of what would be considered a kind of a minimum viable product.

 

Simon Erickson  21:01

Yep. And then one last question, as we’re still talking nerding out on the technologies here Tiernan. I have to ask you, while I’ve got you here, your thoughts on on photonics and optics, because that’s something that wasn’t even being talked about five years ago. And now you’ve got companies like, like PsiQuantum, which Nvidia just participated in one of the funding rounds for, saying, you know, we’re going to look at photons, not electrons, not, you know, atoms or the spins of electrons, completely different way of measuring.

 

Simon Erickson  21:26

It feels super early, but can you tell us a little bit about is this the holy grail of quantum? Is it even feasible to have a photonic quantum computer?

 

Tiernan Ray  21:33

It’s certainly feasible because it’s one of the control mechanisms for an individual particle, you can shine a light on it, and you can manipulate it by putting the individual qubit into whatever kind of state in that bloch sphere you mentioned that you want, and you can get super positions. Here’s the thing, Simon, you can do photonic computing that’s not quantum computing, meaning you can do plain old fashioned ones and zeros where it is not indeterminate. There is no probabilistic element. It is actually a photonic manipulation of a traditional bit. And I think that’s actually more likely to be a near term. Interesting benefit to optical technologies, to all the optical technologies, which is to build chips that are light guides, light wave guides, that are integrated on a semiconductor. And the technology there has also been in development for decades and is getting more mature, and it’s very easy, since you have optical connections in the data center, backing up NVIDIA GPUs in the rack across the room and then actually over the egress across metropolitan areas. You have Ciena and others that are transporting bits now between data centers, because sometimes you need to cluster data centers. You can see the photonic integrated circuit right of PIC being a non quantum plain, old fashioned logical extension of the processing of, say, AI data centers, and so you just speed up using the speed of light, basically. That’s a much more realistic, to my mind, near term viable product that has a clear application. It’s just speeding up either the transmission or the actual processing of the bits in the plain old fashioned real world domain, not worrying about the quantum interactions, yep.

 

Simon Erickson  23:30

And so, just to put a bow on the technology segment of this of this conversation, quantum computers are trying to manipulate particles, whether those be atoms, whether those be electronic spins or whatever else it is, and then, because of the elements of coherence and superposition, just measure them in a way that can do calculations and manipulations of those particles, much, much more significantly, more efficiently than a classical computer.

 

Simon Erickson  23:57

That’s the appeal of what we want to do with quantum computing. Is it correct?

 

Tiernan Ray  24:03

That’s the theory.

 

Simon Erickson  24:04

So let’s talk about how companies might actually be using this now, because we’re starting to see some some science experiments, at least getting some media headlines and coverage from them. AstraZeneca was one of them. They wanted to do a quantum experiment for drug discovery. They set up a designed experiment that you profiled here recently, it was not only using a quantum computer to do it, but it was also using a lot of the GPU cores that were powered by Nvidia.

 

Simon Erickson  24:29

How are GPUs working with quantum computing? And is this, in a strange way, perhaps counterintuitively, good for the GPU industry, that there’s even quantum computers out there.

 

Tiernan Ray  24:39

That’s right, when you do these experiments, such as Ion Q did with the help of Nvidia, Amazon, AWS and AstraZeneca, you can take a problem and you can break it into parts. It might be, in this case, combinatorial chemistry, where you’re trying to speed up all of the work that gets done in the lab at the beginning of drug development to figure. What is the what is the chemistry? What kinds of properties do we want to get out of whatever our eventual drug is? And you have to do a lot of complex simulating of molecules at that outset.

 

Tiernan Ray  25:12

Turns out, in this experiment, a very small portion at the outset was done with, I think, 16 qubits on the Ion Q machine. It was the Forte system. So it’s a current system. And basically you run some stuff in the quantum domain, and then you take you measure it, and you get your normal ones and zeros, and then you put them into the NVIDIA GPUs. It turns out most of the computing work is actually done by the Nvidia GPU, right? So you have a million equivalent cores. When you look at all of the time spent on Amazon AWS, in GPU clusters, it adds up to a million cores running at some I think it was days that they said to do this, so it wasn’t instantaneous. And the reason is the quantum part is a small part of it. I think that one of the quoted speed ups was 600 times. So one element of this problem could be done 600 times faster with the quantum part portion, the Forte portion from Ion Q.

 

Tiernan Ray  26:12

And that’s not insignificant, and it is meaningful. The question becomes, how much money do you invest and how many problems can you find we’re taking the small part is, is worth investing in quantum to do that that several 100 times speed up?

 

Tiernan Ray  26:28

So remember, Simon, the to your earlier point about the theory, what everyone fantasizes about is quantum, quote, unquote “supremacy”. Sometimes it’s called an advantage. It’s really supposed to be for problems that you could never do on a traditional computer. In the case of the AstraZeneca work, you could do this right? And so it might be that you collapse it from, let’s say, months to days meaningful. But then the question is, it, is it a big enough gap that you would not only spend the time to learn how to use the quantum computer invest money in using it at AWS.

 

Tiernan Ray  27:02

But also, remember, invest in millions of Nvidia GPU cores, and at some point it becomes a question of, if you can speed up something that’s huge, is it worth doing that at all? Because you might find other approaches in the traditional computing realm. Maybe they take a little longer, but maybe they’d be more in line with what you’re built to do as an in, as an operation, as a firm, and so you can get these speed ups, but you still are spending tons of money on GPU chips, sure.

 

Simon Erickson  27:36

So there’s a there’s a very small portion of the market. The Astrazenecas are the massive companies that have gotten massively complex problems like drug development. You know how proteins are moving around or interacting with each other for cancer treatments, or what it might be, but certain problems are going to be very time consuming for GPUs to compute. And so perhaps those can be offloaded to the quantum computer. And then it’s an economical question of, you know, is this actually speeding it up enough for us to pursue the quantum versus what’s available?

 

Tiernan Ray  28:09

Right? Exactly. Yep.

 

Simon Erickson  28:11

Okay. And so my second question on this one is, you know, we do have quantum computers available from AWS. You know, AWS Braket has now got a quantum computer. You can actually hired out a dedicated Ion Q quantum computer for $7,000 an hour. That’s more than 300 times more expensive than renting out GPUs on CoreWeave, which is now a big purchaser of Nvidia’s GPU chips, which are $22 an hour.

 

Simon Erickson  28:37

Tiernan, is this still going to be a really complex science project for the largest companies like AstraZeneca? Are you ever going to see, at least, within the next decade, companies tapping into a quantum computer on the cloud to do smaller applications?

 

Tiernan Ray  28:50

You won’t see that, Simon, until you see scaling, which means that you can bring down the cost. You’re very right. However, for an AstraZeneca, $7,000 an hour is nothing, right? You could spend days doing that, and it’d be a fraction of your R&D budget. So you’re right. There are science projects, but for some large firms, that is certainly well within the budget.

 

Tiernan Ray  29:10

So I would expect that there’s a market for AWS Braket quantum services, for large companies that, let’s say it’s, you know, you know, you want to just cover your bets. You want to make sure that you don’t lose track of what quantum can do. And so you devote some amount of your corporate resources in terms of people and money and time to play with these computers on bracket and see what they can do. And you run experiments of the kind that you had with the Forte Ion Q system and millions of NVIDIA GPUs, and you just say, Yeah, we’re going to invest in keeping tabs on it. So yes, there is a very viable market to do that kind of R&D work to make sure you’re on top of it, if you’re a firm that could someday benefit in five or 10 years, if, in fact, scaling happens and the cost collapses. Is, and then it’s as cheap as normal cloud computing. And you say, yeah, it’s going to send all that stuff to the cloud.

 

Simon Erickson  30:07

I want to change gears here and go to a different conversation. Just that is one kind of interesting, you know, other than the Commercial Appeal of, you know, is a quantum computer more economical for a company to offload some of that computing to the quantum machine, rather than the traditional classic. I want to talk if there’s a, if there’s a perhaps this is nationally funded, but a strategic quantum advantage that ever comes into play here, and it’s kind of like an arms race.

 

Simon Erickson  30:33

And, you know, are there certain applications that aren’t just for, you know, AstraZeneca want to develop drugs, but perhaps, if a quantum computer is faster or able to handle more complex things where there are two adversaries kind of going in this chess match.

 

Simon Erickson  30:49

Is there going to be kind of an arms race of who’s got the faster computer that can beat the other one? I’m thinking in terms of cryptography or cyber security protections out there, you know, AI that’s going into military applications, maybe even stuff in the space race. The stuff that makes me kind of scared and shutter even thinking about it

 

Simon Erickson  31:06

But there’s big dollars of contracts from Department of Defense and elsewhere. Do you think quantum has any place in this, or is that still something that’s just kind of a same science project like commercial worlds?

 

Tiernan Ray  31:18

So the common story Simon is cracking RSA encryption that underlies all transactions. And there I cited in the article that I wrote by a consultant, Olivier Ezrati, this being one of the misconceptions, right? And he points out that you’re not going to get cracking of traditional encryption anytime soon. That’s at least a decade away. Companies that are marketing quantum will say that we’re just on the threshold of RSA encryption being completely cracked. However, so that appears to be something of a myth.

 

Tiernan Ray  31:57

However, one of the near term applications that IonQ does talk about, and talked a lot about in their analyst day is, to your point about nation states, that there are a lot of entities, good or bad entities, that are kind of hoovering up existing encryption keys. So they’re taking the keys that underlie current passwords, etc, and they’re storing them in a database, and they’re keeping them there for the day when they get the working quantum machine that will allow them to suddenly crack all of this.

 

Tiernan Ray  32:31

It’s kind of like you take the safe out of the bank and you leave it in your bedroom until you find a drill that will allow you to get into so there’s this idea that stuff is being stockpiled in way of information, conventional information, that will someday be cracked, and that companies are investing in either protecting it by keeping it offline or stealing it to hold on to it for the day. They can break it with quantum.

 

Tiernan Ray  32:57

And one of the near term applications that Ion Q sells is they sell a networking component where there is, you know, cryptically safe transmission via quantum teleportation of these keys, with the idea that if you can’t, you do all the things we’ve been talking about, like efficiently process a full application, you could at least transmit your keys safely encode it as a quantum kind of representation. And so that is a near term quantum application of quantum networking that doesn’t involve any processing. It just involves transmitting bits, usually using fiber optics that would pertain to security, national security, the question of RSA encryption, etc.

 

Tiernan Ray  33:47

And so that’s Ion Q at least makes the the argument that they will get near term revenue from everyone suddenly wanting to kind of safely move stuff around, waiting for the day when, when bad people might be able to crack

 

Simon Erickson  34:03

And implications for the cybersecurity industry, maybe even particularly for cryptocurrencies. You know, we now got digital money that’s out there that’s essentially just code which is being protected by the same security that we’ve had for decades. I mean, is this going to touch the cybersecurity industry in a meaningful way?

 

Tiernan Ray  34:20

It’s not going to immediately, because the cyber security industry can continue to work with the existing RSA encryption that is apparently not breakable by quantum technologies at this point, I think well, before we get to the point where it is crackable, you will have a lot of experimentation by Palo Alto Networks and Fortinet and CheckPoint security vendors with IBM, with industry, where they’re again running these kinds of experiments to see what’s possible. And it goes back to your point about quantum R&D. You have people trying this all the time, trying to run the algorithm. That might crack RSA encryption, and see, are we closer, on a daily basis, to that happening?

 

Simon Erickson  35:07

And then, as I’m kind of closing this out here, and you know, you had mentioned of the publicly traded quantum computing stocks right now, I believe I have it correct that the median price to forward sales ratio of them is 260 times. Right now that’s sounding like the SPAC summer of 2021, super high expectations, even though the very the base is still very, very small. At the same time, we’re seeing hundreds of billions of dollars deployed by the large tech companies into their own data centers, which are, of course, using GPUs from Nvidia and else and others, huge, huge projects, it seems like they have certainly hit scale with this.

 

Simon Erickson  35:46

Do you think quantum computing is investable right now, or is this still something that’s worth monitoring? Could be interesting a couple years out, but you’re not touching anything at 260 times sales.

 

Tiernan Ray  35:55

It’s speculation. It’s not an if it’s trading at 260 times sales, it’s not an investment, meaning you have no support via the appreciation of those assets. What you’re really buying is you’re buying the an option, a call option on one of these companies, Ion Q, Rigetti D, wave will get acquired by a much larger tech company and be added to whatever their stockpile of stuff is so I think most of the stock price currently reflects M&A expectations. There’s always an expectation, if something’s a hot property, it’s going to get bought. And that’s, you know, it’s possible. You do have to consider that. You know, at 260 times sales, these companies are large. I haven’t checked lately the market cap. I’m looking now. So market cap is almost 19 billion for IonQ. Kind of like they’re closing in on 100 million in annual revenue, which is not a lot of revenue, but they have the, you could say they have the most solid income statement. None of these companies make any money.

 

Tiernan Ray  36:56

So to buy $19 billion worth of a company is not a small acquisition. But I guess you could say that somebody out there imagines, yeah, I mean, IBM is going to buy IonQ or Google’s gonna buy somebody’s gonna buy them. That’s what’s gonna happen. So that’s what’s reflecting the price. So it’s a long winded way of saying it is not an investment, because it is not supported by the natural appreciation and time that that can be expected of the assets, it is strictly price momentum based on mostly M&A, maybe to some extent the headlines. As long as headlines keep coming out about the next IonQ system or the next Rigetti system, the next number of qubits, headlines can also add to price momentum. So it’s speculation, sure.

 

Simon Erickson  37:43

And would the quantum computers if we do get to a scale, if we start building this up, do you start to see the IonQs and the pure plays of the world investing heavily in manufacturing quantum computers? Or does that go to someone like a Taiwan Semi, to build some new lines for many I mean, how is the manufacturing you think like this gonna play out.

 

Tiernan Ray  38:00

They insist, some of these companies insist they’ll go and use fully depreciated assets at Taiwan Semi none of them will be building factories. So that’s probably the principal option. It remains to be seen whether they can really use cheaper integrated circuit process technology to build this stuff. I think that’s part of the scaling challenge. It’s probably the case. They will have to go to capital markets again, to your point, do lots more debt and equity raises in order to have the kind of scale manufacturing if they if it’s viable, in order to do it, they’ll need a lot more money,

 

Simon Erickson  38:35

Sure. And then last question is the big tech companies, you know, you always see IBM and Google and even Microsoft kind of, kind of each claiming that they’ve got quantum supremacy over each other. They’ve got more qubits, whatever it is. Do you see them all as kind of equals in this field? Tiernan, is there one that you think has pulled ahead of the others in terms of their innovation for quantum or at least how seriously they’re taking this?

 

Tiernan Ray  38:55

IBM and Google are sort of have dueling positions in terms of the papers they publish. I have the impression with Google’s chip, last December, the willow chip, that Google really had the most clear, thoughtful advances in superconducting qubits. And I would be inclined to give Google the edge over IBM, as far as to heavily publishing quantum research facilities.

 

Tiernan Ray  39:24

However, I will note that others like to point to IBM, the CEO, Nicolo Damacy at IonQ, likes to say that IBM is the only company that is even close to IonQ, which is sort of for marketing purposes. But it’s possible that he sees something I don’t, and that IBM is really in the pole position amongst the giant companies.

 

Tiernan Ray  39:46

But my sense would have been that Google is in the lead, and maybe IBM and Amazon close seconds. And Microsoft’s themajor Rana sort of topological qubit that they’ve been pursuing has had someone. Controversy that I think Microsoft is a far distant third or fourth. It remains to be seen if what they have is actually viable. There’s a lot of scientific controversy about it. That’s my impression from the outside, yes.

 

Simon Erickson  40:17

And then back to the scale question about the, you know, inflection points, you know how this typically works out. We we’ve seen IonQ even in the conversations we’ve been having about them for, I think maybe it’s been five or seven years now, turn and go from less than a million dollars in revenue, all of a sudden they’re at $10 million revenue. Now they’re at $50 million of revenue. It’s a 50x increase in revenue in three years for this company. But it’s still only $50 million I mean, $19 billion valuation. You said it’s highly speculative, but do you think there is enough demand at the top of the big organizations now you’ve got a functional quantum computer, whether or not it’s the perfect one in all things. But is there? Is there going to be a continuation of this inflection point, even from $50 million to 100 to a $500 million and so forth.

 

Tiernan Ray  41:07

Well, I mean, I can tell you what the street thinks. So the income statement projections, you know, are a doubling this year to roughly 91 million. So they’re not at 100 million yet, but they’re getting there. The Street currently calls for, still a very high rate of growth, 87% in calendar, 20 671 million. Maybe they’ll close in if, if the street is right by in four years, they would be closing in on half a billion dollars, and they would be doing so at a clip of 56% now.

 

Tiernan Ray  41:37

I think if they get there, that’s terrific, and would be a validation, because I don’t think they’re going to make it unless they have solved some of the scale issues. There is the scale issue, meaning that estimate is not guaranteed. It’s just an estimate. And until we get there, there’s plenty of opportunity for calendar 26 and 27 to see those numbers also go down. In other words, all of these projections are building on the scale promise of IonQ. So if they can make it to a half a billion dollars good for them, it is, again, still the law of small numbers.

 

Tiernan Ray  42:18

Simon to get to 50 million from 1 million is amazing, but it’s still just $50 million to get to half a billion dollars. You know, Bill Gates used to say when he’s running Microsoft, is all about getting to a billion dollars. And from what I’ve heard from talking with CEOs today, it’s still about getting to a billion dollars. So if you’re at half a billion in annual revenue, your friends and your and your parents will be impressed, but something remains out of your reach, that you haven’t reached the scale of being a billion dollar company.

 

Simon Erickson  42:51

Fantastic. Perpetually five years. That it’s going to be there in five years. And we’ll talk about, well, once again.

 

Simon Erickson  42:56

Tiernan Ray, the technology letter is the the service that he runs. I’ve been following it and a paying subscriber since day one, so I can tell you it is worth checking out if you want to see coverage and innovation, specifically in software and semiconductors and in cool new things like quantum computing. You can also follow him on x at, TiernanRayTech. So follow him there too, some great coverage. And Tiernan, once again, thanks very much for joining me again to talk about quantum here for 7investing.

 

Tiernan Ray  43:27

Thank you, Simon for having me.

 

Simon Erickson  43:28

And a reminder for everyone to do your own diligence. You know, we’re talking about some neat things out there, but certainly quantum computing is still very, very highly speculative. Don’t take anything as the gospel that we talk about on this podcast. We encourage everyone, as an individual investor, to do their own diligence on any positions they’re interested in.

 

Simon Erickson  43:46

My name is Simon Erickson. Once again. Thanks for tuning in to this edition of the 7investing Podcast on quantum computing. Stay safe and we’ll talk to you again next week.