Transcription
Could you explain to us in practical terms how quantum computing differs from traditional computing and what that shift means for the industry?
Quantum computing is super fascinating. It has the capability to do computations that would take hundreds, thousands, maybe even millions of years, even for the biggest supercomputers that are out there today. Now, that's oftentimes in areas like chemistry, material science, and underlying quantum physics itself. It's just impossible to do that.
What is interesting about it is we can actually take that and apply it to do brand new things that would be super hard to do, like creating self-healing materials, figuring out catalysts that would get rid of microplastics, and thinking about how to get rid of forever chemicals. I mean, these are things that we all want to do that would impact all of us positively around the globe. But it's just hard to do that today with any kind of classical computer. So that's basically our goal.
Well, let's talk about how this was able to even be accomplished, because the time horizon from theory to hardware took almost a century. Microsoft's own journey started in 2004. There were some setbacks along the way. A paper was retracted in 2018. Talk to us about that breakthrough that allowed this new quantum chip to actually be possible.
Yeah, this is actually one of the longest-running research projects at Microsoft. I mean, nearly 20 years we've been working on this. And it's because it's a very hard physics problem. So if you look at it, we had to actually go through and prove something that was theorized in 1937, this Majorana fermion, and we had to turn it into a piece of hardware that we could actually turn into a computer.
It has taken that long period of time to do it. But in doing so, we actually figured out a new phase of matter. Everybody learned about liquids, solids, and gas when they were in school. This is a topological phase. It's a new one. The only way to figure it out was to actually create this new thing called the top conductor. It's what comes after a semiconductor.
So it is definitely a high-risk but high-reward solution for it. And that's why it's taken so long to prove out the physics, build the devices, and make it go. But we believe it's the most reliable way that we can get to a million qubits. I can fit them on a chip that fits in the palm of your hand. That's just not something you can do in another way. And that's why we stuck with it this long.
Well, show us what it looks like if you can, and also talk to us about how much it costs.
Yes. So this, and I'll try and hold this up here, but this is actually the Majorana chip that I have here. Try and get the glare off of that. But this actually is the chip. The underlying technology, again, would allow me each qubit, which is the fundamental building block. You think about bits and bytes in the chip in your phone. This is qubits for a quantum computer. Each one fits in about 1/100 of a millimeter.
That's why we're able to figure out how we can get to a million of those on one chip that fits in the palm of your hand. It's a significant amount of work to pull it off. Like I said, we had to have the kind of new state of matter that we had to discover in order to make that work. You know, it's taken a long time for us to be able to figure that out. But, you know, like I said, now that we've got it, we're going to be able to get to a big scale.
Talk to us about what it can actually do. I mean, this is not commercially available just yet, but give us a picture of what the potential looks like.
Yeah, the way I think about this is if you remember back in the 1940s, most computers that we had were actually made on vacuum tubes. At the time, we were always like, well, these are pretty advanced things. Then along came Bell Labs, and they figured out this thing called the transistor and eventually turned it into compute. The industry was born with semiconductors.
We're kind of in that same phase of discovery right now, which is moving from the semiconductor to the top conductor, right? This has this topological phase and allows this new type of compute to come in. Just like at that point, we're at the very early phases. What will happen is we'll get very rapid scale, and now that we've done the physics, we can actually build out the chips. We can make them scale even faster with more power. That will happen year after year after year.
With respect to what we can do with it, like I mentioned, one of the best use cases that we see is going to be in places like chemistry. That's a place where we use even AI now to discover new molecules and build new things. But AI is just an approximation. A quantum computer actually speaks the language of nature, which is quantum mechanics. That's what makes it so powerful.
So we combine those things together, and we're going to be able to do brand new things. Like I said, I'd love to get rid of forever chemicals. I would love to get rid of microplastics. We're going to need to invent new technology to be able to do that. Our computers are going to be able to help us accelerate it. It's just not something you could do on a classic supercomputer.
Let's talk about the competition, because Microsoft isn't the only one developing its quantum computing chip. Google unveiled their own in December called Willow. There's has 105 qubits. Microsoft is harnessing about eight. Talk to us about how we can stack these up against each other. Is it a count, the qubit kind of race, or is Microsoft doing something different in the approach that makes it better?
When you talk about qubits, there is sometimes this race in the industry, saying that they've got a ton of qubits. The real interesting question at the end of the day is what can you do with them? I could have a thousand qubits, but if I can't finish an actual quantum workload application, it really doesn't matter. Or if it takes a thousand years to finish that application, it doesn't matter.
At the end of the day, what's most important is not the count of qubits. In some ways, that's kind of a vanity metric for any vendor, not just Google, but it's true for everyone. What really matters is can I get to that high count but make them useful? That's the value of this topological conductor that we've got, this topological core.
These topological qubits have error resistance built in. They're small, they're fast, they're digitally controlled. There's just no one else that has that. So we can get to the scale, and we can make sure that they're useful. Like I said, today, you're going to start off with a small count, just like, you know, transistors. Originally, there weren't that many of them. There were tons of vacuum tubes. Over time, you're going to get that transition in that scale.
That's what's going to help, you know, and 20 years of experimentation to get us there now starts the real race to go make the volume up and go higher.
So what's the key differentiator between Microsoft and Google's quantum computing chips? How do you know that yours is better than the rest?
I'd say there are a couple of things. One, in the Google example, they announced a logical qubit error detection and correction, but it does not do computation. Of course, for a quantum computer, you need to build compute as well. It turns out that Microsoft actually already has 24 logical error-corrected qubits with the partner Atom Computing, which does neutral atom systems.
That's an example, and that does all the way through to computation. So we already have bigger systems that do computation there today. That's one example, a logical qubit, say, which was part of that announcement.
The next thing is how does one get to scale? A lot of the technologies that companies are using, including Google and others, would need a warehouse of compute to make it work because you can only get so many of those qubits per chip. If you get to a million, you're going to have to have multiple chips in multiple places and network them together.
From a scale perspective, it's going to take a while. There are smart folks working on it. Maybe they'll figure it out. But in my case, if I have the ability to put a million on one chip, I don't have to solve those issues. I can just get the scale with what I've got.
We believe that's actually going to be a more reliable way to get to a million qubits. That's going to be the delta between the two. As a company, we're a platform company. I'm going to keep working with all of the different types of chip manufacturers and the different qubit technology. We want to make sure that they all come through. We have a lot of partners in the space too, so we want them to succeed. We just believe that the solution we've got is the best and most likely to get to that scale.
In January, Jensen Huang said that useful quantum computers are likely 20 years away. Is he wrong?
I think that we're getting to the point, especially with this announcement, where we're years away, not decades away. I believe that part of what Jensen was referring to were some of those scale challenges that I mentioned, that some of these technologies need a warehouse-sized computer to make it work. You don't need that.
If you didn't need that, you can imagine how it would take a long time to get through there. From our perspective, the fact that I can put a million qubits on a chip that fits in the palm of my hand, they're small, they're fast, they're digitally controlled. That means it's years, not decades. There's still a lot of work to do, but it's not going to be decades out. We're actually finding ways to accelerate even as we speak.
Well, even Google said when it released its chip in December that it sees commercially viable applications in five years. Is Microsoft on track to meet that deadline as well?
We've got systems right now, like I said, that already have 24 logical error-corrected qubits, and that is the building block that you use to write applications. That basically started off and doubled every three months last year. We're on an increasing curve, so the next stop is 50 and then 100, etc.
When you get to about 50, then you have a machine that can do more than any kind of classic machine, any kind of supercomputer in any noisy qubit system with just two qubits. That's coming very soon. When you get to 100, you start being able to do science. That's where national laboratories, universities, etc., will do some very deep work that's coming. That's just the next few years.
From that perspective, yes, I do think we're going to see useful computers, and they are going to start showing up over the next few years. It'll start being interesting again for deep research organizations to start off with. As we get more, it'll start to become more something that commercial companies will want to do. Even those commercial companies today, though, are wanting to stay abreast of what's happening, track things, and get their own quantum-ready programs in place. We're encouraging them to do that because you need to be ready when these systems get up and start to scale.
What is the next milestone we should be looking at? Investors should be on the lookout for that tells us that commercially viable quantum computing is around the corner.
I think when you start to see machines that can run applications, not just benchmarks, but applications that really can exceed the capabilities of even the largest supercomputers that we have today, then that's when you're looking at problems now getting solved with quantum that couldn't be solved any other way.
That's going to start kicking in around 50 to 100 logical error-corrected qubits. We're already at 24. Over the next few years, I do expect the industry will be able to overcome those, and they'll start to come through. We're going to just keep going on scale. If you remember your PC days, every year, Moore's Law, I would get double the number of transistors, faster CPUs, and upgrades. We're kind of in that stage now.
But in the quantum space, what's going to be the first sector to benefit from quantum computing?
I really do think that chemistry and materials science is probably going to be the one that gets the most immediate benefit from quantum computing. The problem is, think about why quantum computers are kind of like being able to take an atom and pull it apart and see what's going on inside of there. When that data comes out, then that's something you just can't do in any other way.
Imagine some examples. If I wanted to create self-healing materials, imagine what happens if something gets a crack and it heals itself. Like I mentioned, getting rid of microplastics. We need a catalyst to do that. How do we do the chemistry and the biology to figure out what that is and how to make it work?
We have actual quantum circuits and quantum algorithms today that could do that. We just don't have the machines that would actually run it. That's why we need the scale. Now that we get the machines, as they come online, we'll be able to run those applications, make those discoveries, and have a very good positive impact.
I really do think this area of chemistry, materials science, etc., is going to make a ton of sense. A lot of folks in other industries are looking at how it'll impact them as well. But those are some of the places that we see the most immediate benefit.
Microsoft is also a major player in artificial intelligence. Many of us expected that maybe quantum would come first, given we've been hearing about it for so long. But talk to us about where AI and quantum converge at Microsoft and how it has helped advance some of these quantum efforts along.
Yeah, they're very complementary. I can tell you that they do work in a complementary way, like even writing those circuits that I mentioned. We've actually got components that help people write those quantum applications, you know, just as an aid for people writing apps.
That's a great example. On the other side, being able to do this advanced simulation on a quantum computer is not possible in a supercomputer. We're going to take that data and put it back into our training for our AI. That basically means that our future AI models that we come out with are going to have quantum-trained data inside of them that will make them the most accurate models that we've ever had anywhere at any time.
Now I can start asking it questions, and it's going to give me answers that would have been impossible to do before. So they're very complementary. In the future, you can think of quantum computing as another accelerator. It's going to live in the data center right next to CPUs and GPUs, and the applications that we run are going to be hybrid and use all of the above.
Some investors would say that this artificial intelligence hype cycle has produced more hype than actual reality, at least at the stage that we are now. Is quantum computing going to be different or follow a similar pattern?
I think that, you know, with any new technology that comes out, oftentimes we do spend time trying to figure out what is it good at, what is it not good at. Sometimes we think it's good at one thing and we try it. That's pretty much always been true with disruptive technology.
We're seeing that playing out here. I think quantum has had some of those same kinds of questions. What's it good at? What is it not? As we get up to scale, we will figure that out. We do see places where it's directly applicable, and because of that, we're actually already writing quantum algorithms.
We're already applying it. We've already seen examples where quantum-designed algorithms, even when run on a classic computer, have found breakthroughs, which is fantastic. There's always going to be this opportunity for us to advance the state of the art, and we'll get more mature as the machines get more mature, just in the same way that previous disruptive technology has followed that same path.
We've seen a race emerge between the US and China when it comes to artificial intelligence. Is quantum going to follow a similar trajectory in that way in which we think about national competitiveness? Where does the US really stack up in the quantum landscape?
I think a lot of nations around the world see the value of, obviously, they saw it with AI, but they also see the value of quantum computing for the reasons I mentioned, the problems it can solve. That makes it something that any country investing in advanced discovery is going to have programs on, and they do.
There are smart physicists and folks working on this problem all over the planet. I do think we're going to continue to see that kind of competition. In our particular case, what we're working on is, you know, I have labs both in Europe and in the United States, and our goal is to make sure that we can create the most advanced quantum computing technology.
We believe with the micron, one of the types of conductors that we've invented, that it represents some of the most advanced technology in the space that's here. Now, that gives us a lead. We need to make sure as a country that we continue to invest in that because we have to maintain that competitive edge. You have to expect there's going to be amazing competition, and there are incredibly smart PhDs all over the planet working on this. Luckily, I've got a bunch of them on my team, so we're going to keep pressing very hard.
Help us understand the capital investment that's going to be necessary to make quantum commercially available. Is it going to exceed the sums of money that we've seen already go into artificial intelligence in quantum computing? Is it going to be a little bit different from that perspective?
No one needs a quantum computer in their pocket, all right? But you do have a CPU in your phone, you know, so that's a little bit of a different scenario. I think it's more apt to say that quantum computers are going to be very similar to the supercomputers that came out in the 80s and 90s, talking about Cray supercomputers and things like that, that live in the data center. That's really the place that they're going to go.
Again, they're complementary to CPUs and GPUs in the data centers. My apps will require all of them. From that perspective, I think Microsoft is very well positioned as a hyperscaler because we have these data centers all over the planet. We have all of that compute capacity already, which basically means that I can build out and slot new quantum computers in, and I can basically just link them up with the rest of the fleet.
So we're well positioned for that. That's how you should expect it to go. The one value that I have is the chip that fits in the palm of your hand. I don't need a nuclear power plant to power that thing. I can make it work inside of our existing data centers with minimum power, networking, and cooling, which are things we've already got great solutions for because we're hyperscalers.
I'm expecting it to fit in quite well with Microsoft's portfolio. What's going to be the sort of support that you need from the government for this kind of effort? We've seen with artificial intelligence, there is a need to get that investment, that support from Washington. President Trump and Vice President Pence have really touted future infrastructure projects. Will similar support be needed to really advance quantum computing?
No doubt it will absolutely be helpful, especially as the U.S. looks to maintain a leadership position in the space. In fact, just last week, DARPA, of course, the Advanced Research Project Administration, part of the Department of Defense, announced that Microsoft was one of two companies that were actually advanced into the final phase of their quantum acceleration program.
That was two years of qualified physicists, independent Microsoft physicists, evaluating our data, looking at all of the chip designs and even the devices themselves to decide that this looks like a candidate. They announced that last week that we were selected. We very much appreciate that kind of investment of both time and talent on behalf of the United States. It makes our program better.
We've had nothing but good experiences with the DARPA folks. They've been fantastic. I do think programs like that are important for the U.S. to help maintain a lead in this technology race.