Next in Tech | Ep. 281: Quantum at IBM

View Full Transcript

[00:00:00] Eric Hanselman: Welcome to Next in Tech, an S&P Global podcast where the world of emerging tech lives. I'm your host, Eric Hanselman, chief analyst for industry research at S&P Global. And today we're talking about quantum computing with Ellie Brown, who leads our quantum research. Ellie, welcome back to the podcast.

[00:00:20] Ellie Brown: Thanks for having me back again. It's always a pleasure to be here

[00:00:24] Eric Hanselman: Well, you and I were out at IBM's Quantum Day, and there was a whole bunch that they had tried it out for us to see, the shiny new System 2, but also a whole set of different perspectives about quantum. I guess maybe we can start with a little bit about just your take and what you saw and what you thought the highlights were.

[00:00:47] Ellie Brown: Yeah, for sure. So, it was a fantastic opportunity up at IBM's Yorktown Heights facility. So, they have their research headquarters there with all of the cool nerdy labs, all the way spread out across the building, and we got to see the actively working quantum system too, as you pointed out. Very shiny, very pretty, but the best part was getting to open the front door and see a little bit behind the curtain about what's involved in putting a quantum computer together, all of the different pieces and parts that go into it.

[00:01:19] Ellie Brown: A lot more structure than you usually get on a webinar or something virtual because, you know, it was hands-on, it was very tactile, and you got to see all of the different pieces. You got to hear the fans running on, the classical equipment. It was pretty loud. Classically, we were often informed, was the equipment that was the loud part and that was not acting as well as they wanted it to be. But it was fantastic.

[00:01:45] Eric Hanselman: There was a lot of stainless steel out there, I just, I have to say. This-- I guess we're at this interesting point in the quantum computing journey in which, we've gone from, the lab experiments to now on a certain part of this, there's a marketing and promotion side of this, so it's got to look fancy, it's got to look interesting.

[00:02:07] Eric Hanselman: This really was a prime example of the fact that the classical support, all the sensing, all the other, the gear that was supporting this was really the thing that was burning all the major energy, creating all the noise, and all the other parts of this. But they did reinforce the idea that the amount of energy that's necessary to run a quantum computer is a small fraction of what we typically look at racks and racks of classical computing gear.

[00:02:37] Eric Hanselman: That was certainly an interesting perspective when you start thinking about that. I guess we start to get to those questions of scaling. But they were talking about scale and addressing one of the challenges that we see in superconducting, quantum computing environments, which is you've got this bath of liquid helium there, cooling things down.

[00:02:59] Eric Hanselman: In order to connect all these qubits, you have to have something that's going to be able to link those together. They're talking about now clusters of systems and extending the vacuum to be able to get to more modular capabilities, to be able to get to that cluster because right now linking quantum computers together is tough.

[00:03:20] Ellie Brown: Yeah, definitely. And we've been seeing that kind of in different phases of ideation for the past little while. I think even from the beginning as these quantum systems were coming about, there was a realization that for quantum computers to hit the scale that we want them to, to be very functional, very operable, we've got to have the ability to link them together.

[00:03:43] Ellie Brown: You don't want just one ginormous quantum processing unit that has to sit within an even bigger dilution fridge, you know, and scaling up individually. Nothing that works well today ever works that way, right? You get the biggest gains by linking things together. And so, there's been a lot of focus on how do we do that?

[00:04:05] Ellie Brown: You know, is it through networking? Is it through putting individual chips into these clusters? How do we do that with quantum? Because physics makes that a lot trickier than with classical computers. But it's been exciting over the past year or so to see some of those networking and clustering, ideas move from theory and into practice.

[00:04:27] Ellie Brown: And so as we're getting to these bigger systems and bigger scales, that's what we're seeing more and more, is those clusters, things being linked together, and it just, it makes it operate so much better and in a more, user-friendly actual product way instead of just a lab experiment where you can have a huge system, but, you know, if you're an everyday user, if you want to install it in a data center, you don't want a quantum computer that's the size of a house if you can help it, right?

[00:04:55] Eric Hanselman: Well, I mean, if we go want to get qubit counts up there, maybe we start to get to that point. But as you're pointing out, the real challenge is making sure that you can interconnect all of those sets of qubits in ways that can continue to ensure that you actually can run circuits that are efficient, and that you can actually propagate the kinds of problems that you've got.

[00:05:18] Eric Hanselman: You can run long enough with enough qubits operating on the problem at the same time, which is, again, some of those big challenges. It was interesting to see IBM talk about all of the integration capabilities that they have because, it's something that we're now starting to see, leading to M&A activity with other quantum, vendors.

[00:05:41] Eric Hanselman: The idea that IBM was pointing out that they've got semiconductor capabilities, cutting-edge semiconductors. You've now seen some of the capabilities they were talking about, what they've got in both Poughkeepsie and Albany, and touting that as one of the operational benefits that in fact they're saying that they're able to operate more efficiently because they've now got reset capabilities integrated on the actual semiconductor device that holds the qubits.

[00:06:11] Eric Hanselman: The fact that we're starting to get to a case for price performance metrics for quantum to me says maybe we're starting to get to a point where we're a little more sophisticated. Because they were trotting out numbers about how competitive they were from price performance, in terms of what you could actually do in a particular run, and starting to line up, competitive positioning against other vendors.

[00:06:34] Ellie Brown: Yep. Well, and that's one of the things, that we've been looking at from an analyst perspective is how do you compare all of these different quantum systems? Obviously, you've got different types of qubits, and so you can compare along those broad categories of how do these perform relative to, say a, an ion trap system.

[00:06:52] Ellie Brown: How is that different at a broad scale than our superconducting qubits? But we're really getting into that point where even within these broad swaths of different types of qubits, you have different systems and they have different performance metrics, and how do we determine what good performance is on a quantum computer, right?

[00:07:15] Ellie Brown: There's a lot of debate about what are the specs that you put up as a quantum vendor on your website and say, "This makes my system better than another system." is it physical qubit count? Less and less, right? We're talking about things like fidelity and how long we can keep those qubits running, how good they are.

[00:07:35] Ellie Brown: We're getting into discussions about logical qubit counts, and then as IBM brought up, we have cost for performance, so how much does it actually cost you to run the same problem on all of these different systems? And it's a tricky question to determine the answer to, right? Plenty of different vendors are going to want to put their best foot forward, and so they'll put the metrics that work best for them, which is totally their right.

[00:08:03] Ellie Brown: But it is a different discussion to be having as a quantum ecosystem is moving into these very product type questions and metrics, and as a consumer, how do you actually use this? Where is your value coming from, right?

[00:08:23] Eric Hanselman: Well, in talking about performance, they were also looking ahead to fault tolerance and error correction, which we've gotten into, in a previous episode, where we're talking, it's actually has been one of these ongoing discussions we've had, not only the Alice & Bob folks, QuEra, they're talking about fault tolerance at 2029, and we're pointing out what they identify as the dramatic increase in effective qubit capabilities when you get to that point.

[00:08:56] Ellie Brown: Yeah. So right now, we're in-- there's lots of different names for it, right? The era of quantum utility, I think is one of the most common ways to refer to this sort of in-between era of quantum that we're working in, where we have quantum computers, they aren't lab experiments anymore. They're being used for problems.

[00:09:17] Ellie Brown: Sometimes they perform better than classical systems. Sometimes they perform better along one axis, so they may be faster but less accurate, or they may be more accurate, but maybe not quite as cost-effective, right? Cost is one of those big ones that comes up, quite a bit. And, uptime of the quantum computers, whether they're available.

[00:09:39] Ellie Brown: There's a lot of different metrics that we're tracking right now. But as we move into these later scale quantum systems that achieve fault tolerance, so that means that you can run bigger problems on them. You don't have them breaking a-and shutting down, in the middle of your problem set.

[00:09:59] Ellie Brown: That opens up a different world of possibilities in terms of scale, in terms of what problems you can run on those, and that is achieved, as we've talked about through this error correction. And different roadmaps come at different times, right? We've seen those estimates kind of shrink backwards, which is really encouraging.

[00:10:18] Ellie Brown: It used to be 2030 or later. Now we're seeing 2029, 2028, which is really exciting. We’ll see. Of course, having a fault-tolerant computer doesn't automatically make it business-ready and to scale, but it does open up this different realm of possibility when we're talking about what you can do with quantum computers and the speed at which they'll iteratively scale going forward.

[00:10:44] Ellie Brown: It's the last big engineering problem that needs to be ironed out, and once we hit that, the growth afterwards is expected to be quite substantial.

[00:10:55] Eric Hanselman: I was interested to also hear that they're talking about this rapidly accelerating timeline. They’re talking about new generations of processors that are coming out every eight months, which surprising that they're revving that quickly. Although, again, I guess given the progress they're making, some of the capabilities they're trying to bring to bear, that certainly seems to be a good sign, to my mind in terms of how rapidly they're able to advance some of those capabilities.

[00:11:23] Ellie Brown: But, that just in tune of itself seemed pretty impressive that we're, that we're able to make sufficiently meaningful progress to do something which I'm sure is no small feat, which is to go rev the actual quantum processing devices that you're actually running with, Yeah, no, it was fascinating. And you compare that to some of the classical cycles of replacement, right? In classical data centers and hardware, and the timeline was just so rapid, but very encouraging, right? That we don't have outdated quantum systems floating around for very long. It's like these new improvements are getting implemented across the board and around the globe, which is excellent, I think, for the ecosystem as a whole, not to mention the users that are logging onto the cloud and trying to use these quantum computers as well.

[00:12:09] Eric Hanselman: Yeah, we're still at relatively early days. They were identifying that they had on the order of 90 systems that were currently deployed. Interesting to see that we're still in those kinds of numbers in terms of where we fit. But you'd also talked about different types of qubits.

[00:12:25] Eric Hanselman: They've just made an acquisition with Hughes Research Labs, that is actually bringing on board capabilities of a different type of qubit for them. Spin qubits.

[00:12:34] Ellie Brown: Yeah, definitely. And it was a very interesting acquisition. We've seen across the ecosystem a little bit more of this push away from having one horse in the quantum race, right? We have historically seen quantum companies focus on one singular qubit type, which makes a lot of sense from a research perspective.

[00:12:54] Ellie Brown: You do what you're best at, you pick the qubit that you think is going to be the most readily scalable, the most performant, you either go with the one that you think will have the early gains, but then, you swap to a different one, or you choose the long-term bet that you think will be the it technology.

[00:13:13] Ellie Brown: And that has shifted as a conversation recently towards this idea of, maybe there is no one qubit to rule them all. Maybe we've got, an ecosystem of different qubit technologies, or we hybridize them and we actually put different types together, or we learn from the gains that are being made by one qubit type and we apply it to another, right?

[00:13:38] Ellie Brown: With silicon spin, there's a lot of correlation between superconducting qubits, and there's a lot that can be learned from each type, and they can support one another in that research. So, I know IBM with that acquisition is really focused on not just their current roadmap and their superconducting technology, but looking 10 plus years down the road, what do we want to grow with down the road, not just right now and, immediate scale?

[00:14:08] Ellie Brown: How do we keep this system going far into the future and keep progressing?

[00:14:14] Eric Hanselman: Well, and being able to have multiple horses in that race certainly is helpful. It was interesting to see some of what Hughes has accomplished. I mean, it's a storied research lab, founded by Howard Hughes back in the day. More recently has been something that has been jointly funded by GM and Boeing.

[00:14:33] Eric Hanselman: But this is, as you're pointing out, we're getting to a point at which there may in fact be a plurality of technologies that start to make sense and ways in which you can start to approach this problem. And, maybe we've got, because we've seen with trapped ion, you can get breadth really easily, but processing speed is relatively low.

[00:14:57] Eric Hanselman: Superconducting qubits, you know, wow, much faster in terms of being able to actually run circuits on them, and maybe spin qubits are a means to be able to manage error correction better depending. But yeah, different facets for each of these. So, I guess a lot of different aspects that we've got to keep an eye on in terms of what actually makes an appropriate solution for different applications even. We're starting to get to this point at which maybe quantum is a bunch of things, not just a thing.

[00:15:30] Ellie Brown: I know part of the acquisition was focusing on that idea of manufacturing, too. So, IBM is really focused on their manufacturing capabilities, so not just them being a quantum hardware vendor, but also how do we build these things out, right? And really leveraging their historical expertise.

[00:15:49] Ellie Brown: And with the acquisition, there is that continued focus on how do we bring to bear expertise in supply chains and manufacturing, and making these things actually to scale, not just on the order of 90 systems like we have now, but how do we scale out the build and how do we make it faster, better, quicker, so that when we hit fault-tolerant quantum computers, we'll be able to actually keep up with demand, scale them widely, as widely as, the uptake supports.

[00:16:20] Ellie Brown: And I know that's shifting kind of the conversation as well, because quantum isn't just about the quantum processing unit, right? We've got this whole stack of stuff to focus on, all the way from manufacturing up to algorithms and use cases. And the focus with a lot of quantum companies has been how do we play across the stack, not just in one layer

[00:16:44] Eric Hanselman: Well and that was something that also was a topic of discussion. You start thinking about what else was out there, they were starting to really talk about what is that development environment. How do we get quantum developers? How do we get sufficient quantum scientists? How do we actually get enough people who are smart enough about quantum to be able to keep up with this expanded set of quantum hardware we've got?

[00:17:10] Ellie Brown: Well and I like to think about that problem from two directions. Obviously, there's the education aspect, and you can train up, and you need to train up scientists and users and folks not even within academia, but, within individual companies. So have a quantum team within your enterprise, and they know how to interact with a quantum computer, right?

[00:17:31] Ellie Brown: And make it a viable thing to use as a tool. But then there's also the software side and the build side, and how do we make these systems actually easier for the non-quantum physicists of the world to use? And obviously there's been some very interesting discussions around how AI fits into that, how you can code, VIDE code your way onto quantum systems, and that's really exciting as well right now, is you don't need to even be a fantastic, quantum computing programmer to be able to run circuits on quantum computers right now.

[00:18:07] Eric Hanselman: Well and there also are areas you talk about the intersections of AI. One of the other things that was a topic of discussion was quantum safety, and an area where the intersection of AI and quantum is having some interesting outcomes. I guess it's an area that's near and dear to my heart. We spent a lot of time talking about the importance of getting to a post-quantum, cryptographic, computing environment.

[00:18:35] Eric Hanselman: It was interesting, they also dug into some of the specifics about what things are under imminent threat. And again, if we're thinking about highly capable systems by 2028, are people going to be breaking a lot of the encryption on which the trust of a lot of technology foundation exists today?

[00:18:53] Eric Hanselman: Probably not in 2028, but probably not too soon after. And of course, the problem is, once someone has accomplished this, there's a pretty good likelihood that we won't know it for a while. Because if anybody's able to actually break our current encryption technologies, the most valuable thing would not to be able to let anybody know and actually go be able to decode things, to break certain trust boundaries, those sorts of things.

[00:19:22] Eric Hanselman: But it was, I think, I found it encouraging to hear a lot of their focus on the work that's being done. But the point that I started to make earlier was that now we've got a set of quantum safe, encryption technologies that are out there. AI just did some work in actually identifying that one of the candidates, one of next generation, about 22nd stage candidates that are out there, in fact, was a lot less secure, about half as secure as they actually thought it was.

[00:19:53] Eric Hanselman: So, we're seeing these interesting interactions between the AI side of things, the quantum computing side of things and where they're starting to come together.

[00:20:02] Ellie Brown: Yeah, definitely. And it's moving very rapidly, right? Especially as we're looking at these algorithms. And one thing that was, of course, talked about by IBM, and we've talked about quite a bit as well, is if you're talking security, you really need to think next generation security, which means modular, which means agile, which means adaptable, and not going all in on something because an AI algorithm can come along and poke some holes in what you thought was a really good encryption method, and it's doing its job, right?

[00:20:36] Ellie Brown: I'm so glad that we found that out now instead of going all in on this new method and having to walk it back. But yeah, it focuses even more on that necessity of being adaptable because things are changing very quickly. And I know historically, encryption systems and security as a whole has been I don't know, less than adaptable, right?

[00:20:59] Ellie Brown: It's been very boxed in.

[00:21:01] Eric Hanselman: You're being very kind there.

[00:21:04] Ellie Brown: Um, but you know, shifting your mentality so that you can roll with the threats a little bit better and not have it be this huge undertaking every time you have to change your cryptographic standards, right?

[00:21:20] Eric Hanselman: Yeah, the agility part is really fundamental and, people talk about crypto agility, the ability to be able to go switch encryption ciphers relatively easily. But it's something that we historically, one, we haven't done much, if any of it, in the recent past.

[00:21:39] Eric Hanselman: And two, you're just being able to swap one for another. If there's a lot of work making that change happen, that's not really agility. That's just simply conversion. And we need to ensure, to your point, that we're ready for what is probably going to be a succession of different ciphers, and to have the ability to easily move between them.

[00:22:04] Eric Hanselman: And as I said, not to have this huge effort, get the car up on the lift and take a lot of things apart, but take it off the road in order to actually get it done. We need to make sure that we can actually make these updates with minimal effort while systems are up and operational.

[00:22:21] Eric Hanselman: And, the technology is there. The being able to get to crypto agility is a matter of actually putting the processes and capabilities in place. The technology exists, we've just got to make sure that we actually build it in.

[00:22:33] Ellie Brown: Well and that's the harder part always, right? Is changing legacy systems and changing how you think about things, because inertia is a huge component of being human, I think, and you get used to doing things a certain way.

[00:22:55] Eric Hanselman: Well, uh, curious about your overall takeaway and, has your thinking changed from what we heard? Are there different outlooks? Thoughts about scheduling and timeline shifted? Or are we still on track? Because, what we talked about in the past with our CEO series was that the timeline really does seem to be contracting.

[00:23:11] Eric Hanselman: This is, objects in the mirror may be closer than they appear.

[00:23:16] Ellie Brown: No, it's a good question. I think timeline is interesting because what I'm actually seeing isn't even necessarily a contraction of the timeline, but this idea of focusing on what we're talking about a little bit more, right? So, there is a difference now when we talk about Q-day, it can mean a lot of different things.

[00:23:37] Ellie Brown: You can have quantum advantage, right? Which lots of different companies have claimed to achieve, right? IBM put out their own quantum advantage set of peer-reviewed papers recently talking about, "Here are all of these ways that we have seen quantum computers prove advantageous over classical systems."

[00:23:57] Ellie Brown: So, we have that, which is already in the works right now. It’s 2026 and we're seeing some of these systems that are doing real work. We have the idea of quantum being useful within a business context. While that can mean a lot of different things, it can mean saving money, it can mean saving time.

[00:24:16] Ellie Brown: I think there are companies that are already finding useful ways to integrate quantum into their workflows, and not very many, right? So that's another thing that we can talk about on the timeline, which is wide broad scale quantum adoption, and I don't think that's going to happen in 2028 either, right?

[00:24:34] Ellie Brown: We'll have maybe the emergence of fault-tolerant quantum computers, but it takes a while to get these built up and installed and adoption to happen and even identifying what use cases make sense for these more powerful quantum computers. So that timeline is different as well, and then, as they pointed out at the IBM event, which is the possible breaking of current encryption methods.

[00:25:00] Ellie Brown: That's not going to be the same timeline as business value of quantum computers. That might be a little bit further down the road, and we might not know when that happens. So, in some ways, the timeline is coming closer, but I think it's also becoming more clear. We have a lot more clarity around when different things are going to happen, and I think that's very encouraging over the next few years to look at the timeline and say, "Okay, we might see this, but we probably won't see this."

[00:25:28] Ellie Brown: And I think that clarity is probably the most useful thing, beyond even compression of the timeline.

[00:25:34] Eric Hanselman: Well, that clarity is certainly useful. But I’ll point our listeners to our research and some of the takes on some of this. But a lot more to keep track of as we move forward.

[00:25:44] Ellie Brown: Yes, that's true. And if we're talking timelines and research, we've put out a really good, if I say so myself, a nice report covering some of the different roadmaps that are coming out over the next couple of years. So that would be a great one for people to read up on and get a nice visualization of the timeline.

[00:26:02] Ellie Brown: It's always hard on a podcast to speak out a timeline into existence. So, if you want a nice visual, that's a better way to go.

[00:26:11] Eric Hanselman: Yeah, that's awesome. Well, this has been great, Ellie. Thanks very much, but that is it for this episode of Next in Tech. Thanks to our audience for staying with us, and thanks to our production team, including Sophie Carr, Feranmi Adeoshun and Dylan Scheible on the marketing events teams. If you liked this episode, please subscribe or like us. I hope you'll join us for our next episode, because there is always something next in tech.