Bloomberg Audio Studios, Podcasts, radio News.
Yeah, welcome to this special edition of Bloomberg Technology. I am Ed Ludlow. We're live in San Jose and in videos GtC conference, where the major focus today has been quantum computing. Over the next thirty minutes or so, we're going to speak to Nvidia's senior director of Quantum, Tim Costa, as well as the leaders of quantum computing companies ion Q and d Wave. And actually I want to look at the shares of those companies. There was some downward
pressure on a number of quantum computing stocks today. I flagged that only because that pressure continued as they were speaking on stage alongside Nvidia's CEO, Jensen Jog. Now, remember quantum stocks went into free fall on January eighth, after in Vidia's CEO said we were more than a decade away from quantum computers being able to do anything useful. Now, fast forward to today, and Juang had this to say about those earlier comments. Listen, I'm a.
Public company CEO, and every so often someone asked me a question, and most of the time, most of the time, well some of the time, I'm trying to lower the bar here. Some of the time I say something right and sometimes sometimes it comes out wrong.
Now here's the thing in the world of technology. Historically, at least, these have been two distinct fields, quantum computing and accelerated computing or supercomputers for AI, but increasingly those worlds are coming together. Joining me now is Tim Costa, who's the senior director for quantum Computing Adam Vidia, and I think a really important place to start is what
Nvidia does and what Nvidia does not do. Sure, Nvidia does not make quantum computers, nor does it sell quantum computers, but it does provide architecture software, and so this is to that industry. Just explain your role, please. Yeah.
So you nailed it on the head with the easy part, which is what we don't do. We do not build a quantum quantum computer, but we have a vision about how quantum computing will be useful. And it's really a point at which we have the integration of larger scale quantum technologies quantum processors as part of data centers and
large scale computing. It looks very similar to what we have today right these If you look at today's large scale computing infrastructure, it's CPUs and GPUs and storage and memory and interconnects, and it's very complex, but better heterogeneous, and each part plays the role that it's best suited for, including the CPU and including the GPU. Quantum technology offers promise to be very good at certain kinds of computation, and so that will be an additional element of that
system and come in. And so what we're focused on is really helping a helping the quantum tech companies who are building those technologies to better develop those technologies because we're interested in solving the problems that it will be able to solve, but also be setting up that infrastructure to be the best partner to that quantum device right to be able to do things like error correction, calibration
of the devices. These are in some ways things you can think about as physics experiments upon which we try to compute as much as a computer, and so managing that physics experiment and getting the right results out of it is actually a complex computational task that we look to our accelerated supercomputers to actually perform.
What we are literally talking about here is a computer scientist or engineer at a quantum computing company with access to some of your GPUs in whatever form. Fact, it could just be the single GPU could be scaled up to that server design, which we know as DGX. What's the market there? You know? How widely is your technology being used in parallel with many of the quantum names that we saw today.
Yeah, So what who we saw on stage today was a great selection of our partners who are all engaged with and who all are using GPU technology as well as accelerated computing technology including the force acts and other components that we develop in their research program to simulate their devices and build better versions of their QPUs, to work with their clients and work on algorithm design by simulating a quantum computer, and also doing the fundamental research
to drive towards that vision that I just discussed a minute ago, where we actually have this tight integration of these devices together. That involves work on the interconnect between the QPU and the GPU, that involves developing new methods for error correction that can be deployed at scale on a large supercomputer using novel AI methods, among many other areas. But those are not our only partners in this ecosystem.
We're working with over one hundred and sixty groups in quantum computing, and the range of application areas is quite wide. But they're all using in video technology to accelerate their work because that's what we're ultimately here to do.
So we started the day with the idea that in vidious technology can help quantum computers accelerate their own timeline reach that useful metric. But the idea raised and put Jensen Wong by those partners was actually the output of a quantum computer. In other words, the computation can work both ways. It could be submitted for use in the training of a foundation model. What do you make of that?
Yeah, So I think that there's two sides of AI and quantum and they're both really important and really fascinating. And what I touched on a few minutes ago was the AI four quantum right, using AI models to actually control an error correct larger and larger and more capable quantum devices. It's incredibly important and it pulls in the timeline useful quantum computing.
Yes.
Now, the other side that was brought up today, and it's a fascinating topic, is you know, what is a quantum computer? If you start to boil it down and I won't try to go too far down, but it's they're really physics experiments. As I said, you're modeling quantum physics in the device, and so they're able to potentially provide data to train and fine tune models for understanding the very phenomena which are inside of a quantum computer in a way which will answer questions that humanity is
and they've unable to answer. So we think that that's a really interesting and exciting area to pursue, and we're engaged with our partners across all these different areas.
I don't think that in the context of all the attendees I've spoken to you today and Jensen and the panelist, that we've kind of reached definitive agreement and what useful is. But I think we've definitely reached agreement. But within your industry, that this accelerates what's happening pardon the pun, accelerated quantum computing. What happens next for Nvidia and their footprinting quantum. There
is going to be a research center in Boston. That those can often be abstract things, But why is that an important step?
It's an important step because we're going one of the things that we have to do if we're going to build a new kind of computer, and if we add a quantum accelerator to a computer like what we build today, that is a new kind of computer. You're adding a new computer ELEMENTUS. That is a physical endeavor. It has a footprint, needs a place to do it.
Now.
The center in Boston won't be the only place that that happens in the world, but it's a place for some of our partners and and US can work on developing the interconnect, developing the air correction technologies literally string up their quantum devices to our GPUs and build the first versions of these quantum accelerated supercomputers that we're all working towards.
You'll view on what is useful. What do you think that a quantum computer will be able to achieve, whether it's assisted by nvideo or not sure.
I think that one of the most important things that Jensen talked about today on stage was really narrowing the focus and deciding what the one the problem is so that you can define success and chase after it. I do think that there's fairly wide agreement in the community that one of the first areas is to be to be accelerated and to have new kinds of problems solved that weren't before. Is in chemistry, biochemistry related areas. I mean, there's some kind of sniff test. This passes.
Right.
You've got basically quantum physics in the quantum device, and the ability for that to model quantum physics in terms of what's required for very accurate chemistry just kind of makes sense. So we think that that's going to be the first area that's a disrupted early side you. I'm sure there's people on my team in discree me as we start today from the panelist. There are a wide variety of opinions on everything in quantum, but we think that's promise them.
Well, we're grateful for yours. Tim Coster, Senior director of Quantum Computing at Nvidia, thank you very much. A lot more coming up. We speak with ion Q executive chairman Peter Chapman. That's next. We'll be right back. This is Bloomberg Technology. Welcome back to a special Bloomberg Technology at nvidia's GtC Quantum Day. Ion Q was one of the quantum computing companies invited on stage today alongside Jensen Wang
here at gtc's Quantum Day. The company's stock was one of those hardest hit January eighth, after Invidias Jensen Wong made those comments about quantum computing and its usefulness being quote a decade or more away. Since then, ARMQ has gone through some changes, appointing Nicolo Demasi as its new CEO. The former CEO Peter Chapman continues as executive chair and
I'm delighted to say joins us now on set. And what was interesting in the conversation with Jensen Wang is that you were balanced in saying I'm not I don't think we necessarily agree on everything here. Still, let's start with the main point, which is did we define usefulness and do you have a sense that Jensen Wang has changed his timeline of when he thinks usefulness of your industry will be achieved?
Well, I think today was kind of the purpose of today was to bring that timeline in to kind of take back what it is that he had said before. He said on stage I think twice it was his mea copper yes, right, So that was the purpose of today's kind of.
He said that he would be the first CEO probably in history to invite a panel of people to te them that he was wrong.
It was wrong, is exactly so, and it was funny today. Just in general, we've been working with Nvidia on a demonstration for today that was with Nvidia AWS and Astrozeneca, where we'd gotten a twenty x improvement on what we had done previously. Also, we had taken with answers with a product that they do which is normally run on GPUs, and managed to get a twelve percent increase in using
our quantum computers. So while those numbers are not really enough to take over the market, because usually you need one or two orders of magnitude to be disruptive to a market, the fact that we managed to do it on the kind of our thirty six cubit system is really quite remarkable.
Nvidia has a quantum computing business in so far as we got into it during the panel. That you have access to architecture, DGX, the hype performance, GPUs, also some software and open source solutions, and what they would say is that that will help your engineers and computer scientists calibrate, reduce err accounts, design better. Is that how it actually plays out for you some aspects.
We have a DGX cluster that we use for a design for designing other quantum computers, not so much in the error correction aspect. That's something we do ourselves, but certainly we use a number of GPUs for designing the quantum computer itself. We also for small cube accounts, because soon as you get into me on about thirty five cubits, you can no longer simulate one of these on a GPU, So for those we actually run the simulation on a GPU just to make sure our hardware is working correctly.
The problem is, when you get to sixty four cubits, you need two and a half billion GPUs.
Because for each cubit you add to perform its essentially double bubbles.
Right, and it means that the matrix math that you have to do suddenly is doubling as well. So basically is it about thirty five cubas To fully simulate it, you can only get it on a single dgx one hundred.
I got to hold you to this, and I wonder if it ties into the new CEO and you know, sort of reset a little bit. There's a difference between lab experiment and commercial use, you know, making money, revenue generation. That's the question I get for you most.
Yes, so what Jensen said, we actually one area we definitely agree on, which is you need to find a set of applications early on that you can start to make money on and to build that firewheel to be able to power your R and D. And so the examples we're talking today, for instance in the chemistry application and also with answers is exactly those kinds of things. So that's exactly what our plan is.
PCU or the executive chairman. Now you were CEO one week ago, Kerisdale issued a short report on your stock on your company. I just want to give you a chance to respond, as we've not had a chance to speak since then.
Sure, it's a short report, so their goal is obviously to try to cast out about the company and make money that way. Unfortunately, it's just a if you're a public company, these are the kinds of things you have to endure. You know, we don't put much credence to those things.
The last thing I want to ask you is about how big today was in the change of trajectory or momentum for your industry. Jensen Wang is a character and he was honest on stage. But GtC is an incredible event. It has scale, it has eyeballs. Do you think you'll see something of substance come out of this for your company, for your sector?
You know, it's an interesting it's certainly Jensen's goal was to give us the microphone today to be able to get out our story, and that's certainly important. But I say, if you were Sam Altman five years ago trying to convince the world that AI is coming, probably he wouldn't have been successful. And the question is is how much time should Sam Altman try to convince the world AI is coming? Instead just go back and actually make it happen.
So in that sense, actually it's probably not that significant. What really matters is actually going and doing it, not actually getting the message out.
I'm Q Executive Chairman Peter Chapman. Thank you for your time here in San Jose at GtC. Okay, much more to come. Alan Barrat's CEO d Wave, another one of the quantum computing CEOs on stage, joins us, and they have an example of success or something useful blockchain architecture progress. That's next. This is Bloomberg Technology. Welcome back to a
special edition of Bloomberg Technology Live in Videos. GtC Quantum Day so shares of some quantum computing companies sank today after leaders spoke at an event with in video CEO Jensen Huang. The plunge follows a similar move back in January after Huang said it would be more than a decade before the technology quantum computing could do something that's useful. One of those quickest to say in the month of January that Wang was wrong was d Wave CEO Alan Barrats.
D Wave shares under pressure today for whatever reason. But Alan, I'm grateful for your time, and I think it's fair to say that among the panelists you maintained those areas that you don't agree with Jensen Hwang on. How in any way is your mind changed on those differences through the course of today.
Well, my mind has not changed. The fact of the matter is that we at d Wave have taken a very different approach to quantum computing from everybody else in the industry, and as a result of that, we are actually able to support useful, important applications today.
And I think the best.
Example of that is the paper that we published in Science last week, where we have demonstrated that we can compute properties of magnetic materials that just cannot be computed classically, and this gives us the opportunity to create new materials discovery platforms which will dramatically reduce the time and cost to create new materials.
And that seems pretty useful to me. Yeah, there's also some applications that the audience might find harder to understand. For example, blockchain architecture. Why is a quantum computer able to improve that process where a supercomputer classically coded in ones and zeros cannot.
So basically, what we did was we were able to show that that same computation that we use to compute properties of magnetic materials could be used to basicly we compute the hashing functions that are used in blockchain and cryptocurrency. What this means is that we can now create a blockchain that uses a quantum computer to do the proof
of work. What's so important about that is that quantum computers consume far less energy than classical computers, So this means cryptocurrency mining could be at a fraction of the energy cost of what we're seeing today.
One of those two case studies revenue generates this VIE.
Well, the first prototype of this blockchain is running right now. We have it running on four of our quantum computers. It's the first distributed quantum application where each of the quantum computers can create hashes or validate hashes, basically run the proof of work algorithm. And we're now in the process of building that out so we can get to the point where we can support a full commercial blockchain. How long do I think that'll take? Yeah, I think we're looking at a year or two.
Not the ten or fifteen. The ten or fifteen or twenty two. So there is work than a video.
But by this, I mean that's just one application, right, I mean the materials discovery that's running today. We have customers like Entity DoCoMo. They're using us today for cell tower resource optimization. So we are useful today. Blockchain is one of our newest application areas, and yes, we're just starting to roll there.
The point of difference I think, I still think is Jensen's definition of usefulness. Perhaps, But you know, Nvidia does do work with your industry. To summarize, it's basically offering GPU access on the architecture side, as well as some research and open source facilities on the other side. And their argument is you can take that and use it to make your quantum computers better calibration, air account reduction, and I think other your colleagues mentioned design, how do you work with video?
Okay, so annealing quantum computers do not have the same error correction requirements as gate model quantum computers. So we are solving useful problems today without error correction, and as a result, that component of what Nvidia brings to the table is not all that important to us today. When it comes to calibration. We have the largest quantum computers in the world. Our current systems are at five thousand
cubits and growing. We calibrate them ourselves. We don't need GPU paler to calibrate those systems, so currently we do not I mean, I know Jensen says he works with all the quantum computing companies, but d Wave is quite different. We are a different We've taken a different approach. We are at a different level of maturity, much more mature than the other quantum computing companies. We are delivering useful applications and useful value today. Now that having been said,
we are also developing a gate model quantum computer. The approach that everybody else has taken for that effort, we will be looking to leverage some of the same sorts of things that the other quantum computing companies are leveraging. But we view a kneeling and gate as very complementary. They solve different classes of problems.
We only have thirty seconds. It was a pretty public disagreement, argument debate. Will it help you in the long run? What happened today here in San Jose.
I don't think this event was all that helpful to the industry or to d WAVE.
I think.
I thank Jensen for the opportunity to participate. I think that it was great to have the opportunity to try to get the message out. But I think we're still at the beginning of a learning curve with respect to how Nvidia and Jensen interact with quantum computing companies.
D Wave CEO Alan Barratz, thank you for your time here in San Jose. Whether you agree with your host or not, Well, that does it for this special edition of Bloomberg Technology a lot to re cap, particularly when it comes to quantum computing, So don't forget our podcast. You can find it on the Bloomberg terminal as well as online on platforms like Apple Spotify, and iHeart Live from San Jose, California, at GtC and Video's Quantum Day. This is Bloomberg Technology