← All episodes
Wil Oxford

Episode · Dec 20, 2022

Wil Oxford, Anametric

Why QKD fell out of favor and random numbers took its place

Summary

Wil Oxford, founder and CEO of Anametric, a company developing new technologies for chip-scale quantum photonics, is interviewed by Yuval Boger. Wil and Yuval spoke about Anametric’s initial cybersecurity focus, why QKD is not a popular word in Washington DC, why the US Air Force is interested in Anametric’s products, and much more.

Key takeaways

  • Anametric started out building chip-scale quantum photonics for QKD but shifted its commercial focus toward quantum random number generation.
  • QKD carries baggage in Washington DC policy circles, making it a difficult word to lead with when pitching government customers.
  • The US Air Force has shown interest in Anametric's photonic chips, tying the technology to concrete cybersecurity needs rather than abstract quantum hype.
  • The pivot reflects a broader lesson in quantum photonics: finding a product that customers will actually buy today matters more than chasing the most futuristic use case.

Read the full transcript on The Quantum Computing Report here

Transcript

Yuval Boger: Hello, Wil, and thank you for joining me today.

Wil Oxford: Thanks for having me.

Yuval: So who are you and what do you do?

Wil: I'm a former architect at Apple Computer turned serial entrepreneur. I'm a bit of an anomaly. I love the research side of things, and I'm very academically inclined, but I'm a devout believer in the whole practice of product engineering. So that puts me, in some ways, with a foot on both sides of what is usually a pretty wide divide. Currently I'm leading a very talented team that's pushing the envelope in building truly scalable and affordable quantum devices.

Yuval: And these are quantum communication devices, quantum security devices, quantum computers? Which kind of quantum devices are these?

Wil: Well, our initial market is aimed squarely at the security space. I've been in the cybersecurity business for probably the better part of 15 years now, and a lot of my cohorts at Anametric are also very experienced in the cybersecurity field.

Having said that, what we are building is a foundation, and we're building a foundation of, like I said, small, scalable, and affordable quantum devices. In this particular case, that really means that they're quantum photonics devices. So we're building things in the quantum photonics realm, and the first application area that we believe is very clearly needed is in the security space.

Yuval: Okay. So what does the device do, to the extent you can tell me?

Wil: All right. So our initial product is based around a quantum core, a quantum random number generator, and it's photonic, as I'd said. I could go into a long description of why quantum RNGs are in fact necessary, but I'll just say that that core allows us to then build layers on top of it that allow us to do post-quantum security in a more performant and, frankly, more secure way than a lot of other implementations.

Yuval: Where does the photonics come into play? Because quantum random number generators, you could do this as a service. You could have chips that have maybe non-photonic elements that generate random numbers. So where does the photonics come into play?

Wil: Sure. So the interesting thing about photonics is that photons are not affected by nearly as many external influences as other quantum effects.

So if you look at a quantum diode, basically you're talking about electrons tunneling across a junction. If you put a strong magnetic field, for example, around that, then you're going to bias that action quite heavily. If you put a strong enough electrostatic field, the same thing will happen.

If you put a really strong magnet next to a photonic emitter, it just ignores it. And so in a lot of ways, it's a more secure, more robust, and in fact lower-power way of generating quantum randomness or quantum entropy. So that's one of the reasons for photonics.

I think the other part of it, though, is that if you look at quantum random number generators, one of the things you really want to do is to be able to prove that it is, A, operating as you intended, so it's not generating a bunch of strings that repeat each other or repeat itself.

But you also want to have some path toward a computational quantum element, and that's another thing that photonics really allows us to do in a more sophisticated way than if you just had a quantum diode, which is useful for generating entropy, but it's not a structure that you could use for computation in the future.

Yuval: So you're describing a device where the output is also optical. So it's not just optics and light on the inside, but it also emits light. Is that correct?

Wil: That's right. So we do generate single photons in our chips, and we also detect those single photons in our chips. So ultimately, the input and output are in fact electrical, but the mechanism, the quantum effect, is completely photonic in nature.

So when we generate those single photons, we then put them into superposition using a Hadamard gate. Our Hadamard is an order of magnitude, if not more, smaller than every other photonic Hadamard on the planet. And then we do some other processing on the photons themselves to make sure that we get the right kind of distribution, for example.

So these are all photonic mechanisms. And yes, they are controlled by electronic signals, and ultimately, when we read out the output of the quantum RNG, it is in fact an electrical signal. But at that point it's a digital signal and much less susceptible to those kinds of interference.

Yuval: At what stage is the product, and who would be the customers, or who are the customers?

Wil: So we're still relatively early on. We've had four generations of chips already in the lab for testing, and we've just taped out two more chips that will be back in the lab early next year. We expect to have a fully functional prototype device in the middle of next year, so the middle of 2023.

Currently our largest customer is the US Air Force, and we have multi-year contracts with them already underway. We've already completed one two-year contract with them. I think that they are ultimately interested in the longer term, but they're probably not our shorter-term, high-return kind of customers.

So we are still in the process of working to come up with a small number of commercial enterprises that we can sample our devices to. And like I said, we're expecting to do that in the middle of next year.

Yuval: And would these be phone manufacturers? Would these be creators of security products for networking? Can you give me any sense of that?

Wil: Yeah. So our initial group of customers is in the secure networking space. Basically, if you look at any kind of encryptor device or a VPN, you're looking to encrypt a transmission so that someone can't tap into it. And as I mentioned earlier, there's a post-quantum crypto transition that's happening in this space, and we're just at the very beginning of that.

Of course, NIST still hasn't finalized it. They've come up with the final set of standards, but they haven't issued a standard yet. And so there's still some room for innovation, if you will, in that space. But having said that, our quantum RNG is designed specifically to feed a post-quantum crypto implementation, an endpoint for that.

And so we believe that our implementation of a PQC accelerator, all on one chip, with a quantum photonic device at its core for RNG and other features, will in fact be an industry leader on the day that it's announced.

Yuval: I think I read someplace that the technology is based on research that was done at SMU.

Wil: That's right. Yeah, so SMU's been a great partner for us from the beginning, actually from the pre-Anametric days. I mentioned I've been in the crypto business for a while. I had previously started a company called Rubicon Labs back in 2007, and as part of that company, I had gone to SMU to basically figure out how I could use their supercomputer, which was sitting around running calibration cycles, frankly.

That's a long story, but the point of it is, I met the chairman of the computer science department there. We got along famously, and they completed a great two-year project with us at Rubicon. So at the end of that project, I looked at Mitch—Mitch Thornton is the professor—and said, "You know, Mitch, I really like talking to you, so let's just get together once a month and talk and see what happens."

And so we did that, and at the end of one of those monthly conversations, we'd filled a whiteboard, and Mitch looked at me and looked at the other guy in the room and said, "Well, I think it's time for you to go form another company." And that was actually the genesis of Anametric. So I spun Anametric out of Rubicon Labs in 2017, and Mitch and his team have been a great partner for us from day one. It's really a wonderful relationship.

Yuval: I'm trying to connect the dots. You mentioned that there's going to be a post-quantum crypto accelerator chip that's based on photonics, and then you mentioned that the Air Force is actually interested in something that's going to happen not right now, but in the future. Is that the same thing, or is this just the first layer of the foundation that you were mentioning earlier?

Wil: Well, it's a little of both. But I will say that the Air Force is very active in the three major quantum technology areas: photonics, certainly, but also the transmon, or superconducting qubit, area, as well as ion traps.

And so they have their toes in all those pools. We are working with their photonics group, not so surprisingly. But in general, what they're looking for is the same post-quantum transition that everybody else is looking for in the short term. And the simple matter is, if you put a satellite up in geosync orbit, it's $100 million just to get the satellite up to that orbit.

Forget about what the cost of the actual satellite is. So if you're going to spend that kind of money, you want it to have a relatively long lifespan, and by that I mean 30 years plus. And if you look at the various people you've had on your show who talk about the post-quantum threshold, when it's going to be absolutely a requirement, that estimate keeps coming down, and now the latest estimates are, oh wow, it could be 10 years.

And so if you put a $100 million satellite up in the air and 10 years later it's completely hacked, then it's probably not a good idea. So if nothing else, good use of taxpayer dollars insists that you must look at post-quantum crypto as a foundational element of future satellites.

So that's the nearer term for the Air Force. As you suspected, there are other, longer-term things, and one of them is quantum networking. That's another area that we will be active in, more so than we are now. Right now we're focused solely on getting a product out the door and generating revenue. That's really our nearest-term focus.

But longer term, quantum networking is a really big part of where we expect to play. It's really just a very small step from quantum RNG to taking those photons and sending them off chip, or bringing external photons on chip, to do the same kinds of processing that we're already talking about.

So quantum networking, whether you want to talk QKD, which is sort of a no-no word in DC these days, or entanglement distribution, which is a lot more popular—either way, you still need the kinds of structures that we've got on our chip, and so that's our next area of concentration.

And then ultimately, of course, everybody's interested in the quantum computing aspects. We are as well, but we think of that as a much, much longer-term perspective for our little company. Of course, there are others who are looking at it right now.

Yuval: You've made me curious. Why is QKD a no-no word in DC?

Wil: Well, it's a long story, and I won't get into all the exact names and ranks of the people involved, but basically, oh gosh, probably four years ago—and I could have the date wrong—the NSA sent out a memo saying, "Hey guys, we find QKD to be non-interesting, and here's why."

And it gave five reasons, basically, and some of those reasons are absolutely still true today. Some of those reasons are maybe not so true. And the effect was that it put a chill on QKD research in this country and on QKD commercialization efforts in this country.

And that certainly has had an impact on people coming up through academia and saying, "Oh, if I'm going to get out of grad school, and I've got all these student loans to pay off, I need to find a real job, and it's not going to be in a company that's doing QKD."

Now, that was years ago, and things have changed a little bit. But having said that, I think it's good to note that the NSA came back out and said, "Look, we're not saying that quantum communication is bad. We're just saying we've noted these problems with QKD." So entanglement distribution is definitely a good word, if you will, in DC.

I'll finish up by saying two more things. In the EU, of course, it's almost the other way around. The EU has definitely been all over QKD, and they are actively involved. And China, of course. The interesting thing about China is that they put up a satellite, the Micius satellite, a few years ago, and that was immediately called a Sputnik moment for the US quantum industry.

And so whether they intended to or not, China basically boosted US quantum awareness.

Yuval: So tying together the revenue that you spoke about earlier, DC, Europe, and China: do you expect to be able to sell your chip outside the US, or do you think there'll be some strict export restrictions on it?

Wil: We've had discussions with people in the export control area, and I think we're all in agreement that there are certain things where it's just better if we have open exports. Now, having said that, there are certain applications where clearly that's not the case. So what we've done with our particular chip, for example, is we've separated the execution environment, if you will, into two areas.

So there's a very deep-rooted sandbox in our products. The low-level quantum part runs code that we write and that we're basically responsible for, and then on top of that sits an application layer. And the application layer is really up to the customer to decide exactly what it is that they want to run.

Now, we'll provide our own example applications if our customers want them, and one of those applications is PQC key exchange. But that doesn't mean that's the only thing you can do with this part. And so some of our classified customers, for example, want to run their own code on our chip that we have no visibility into, and it's classified, and that's just fine with us.

We take suggestions from them as to what they would like functionality-wise, but we don't have any visibility into the actual business of what they're doing, and that makes a very clean export control division. So we can sell our chip with completely open, even open-sourced, applications that anybody can use or customize however they want.

But the things that are classified are typically on the application side, and we don't have any visibility into that.

Yuval: You mentioned it's a small chip, but my question is, do you need a large company to build a small chip? I mean, how many people are you? How large do you need to become to complete that first product?

Wil: Right. So we're actually very small. We're less than a dozen people overall. But I'd say a good third of us are highly experienced in building silicon. I started at Apple, gosh, 40 years ago, 35 years ago, and I spent almost 15 years at Apple building chips.

At one point there were about a dozen or so years when every Macintosh that went out the door had at least one, if not two, of my chips in it. And so even though we are small, we understand what it takes to build chips. And as I mentioned earlier, we are already on our fourth-generation part today, and we're working to close the gap between tape-outs even further.

So yes, we're small, but we are taking advantage of some excellent partners in the silicon space. We're using the AIM Photonics foundry up in Albany, New York. This is a foundry that was set up by several groups together, but it is a state-of-the-art, 300-millimeter fab, and in the same fab people build, for example, five-nanometer designs.

We're not pushing the envelope nearly as much, but the process technology is the same and the equipment's the same. And so we've been really lucky to partner with AIM, and it's been a great partnership for us for at least three years, maybe longer now. We couldn't be happier with that.

Eventually, we will move our design over to a more commercial fab like GlobalFoundries, or one of several options there. And the great thing about that is we're designing our chip to be able to be built anywhere. We're using 65-nanometer design rules, which are ancient, but what that means is that you can build them pretty much anywhere in the world.

And we're trying our best to make sure that our technology is scalable, not just in the size of the design, but also in the process technology. So we can scale down to 45 nanometers, which is one of the more well-known foundry technologies around, and maybe 22 without much effort.

Scaling down to five just doesn't make sense for us. It's too expensive to build masks for chips these days, unless you need three billion transistors on a chip.

Yuval: So as we get close to the end of our conversation, I wanted to ask you about dinner. And my question about dinner is, if you could have dinner with someone in quantum, dead or alive, who would that person be?

Wil: Oh, no question: Anton Zeilinger. And this decision of mine was long before he just got the Nobel Prize, right? I find him a fascinating character. I've watched a lot of his interviews. I've seen some of the specials that have been done on him. I just find him a most incredibly fascinating person, so no question.

Yuval: Fantastic. Well, Wil, how can people get in touch with you to learn more about Anametric and your other work?

Wil: Well, we are still relatively in stealth mode. By that I mean we're open about what we're working on, but specific details are harder to come by. We will open the kimono a little wider in the middle of next year.

If you want, you can go to anametric.com, A-N-A-M-E-T-R-I-C, and you'll see some very generic stuff up there about what we're working on. And if you're interested in learning more, you can send an email to info@anametric.com, and it'll get to me. We'll be glad to continue the conversation as appropriate.

Yuval: Excellent. Well, thank you so much for joining me today.

Wil: Thanks for having me.