Nobel Laureate John Martinis Discusses Superconducting Qubits and Qolab

This episode is a first for the show - a repeat of a previously posted interview on The New Quantum Era podcast! I think you'll agree the reason for the repeat is a great one - this episode, recorded at the APS Global Summit in March, features a conversation John Martinis, co-founder and CTO of QoLab and newly minted Nobel Laureate! Last week the Royal Swedish Academy of Sciences made an announcement that John would share the 2025 Nobel Prize for Physics with John Clarke and Michel Devoret “for the discovery of macroscopic quantum mechanical tunnelling and energy quantisation in an electric circuit.” It should come as no surprise that John and I talked about macroscopic quantum mechanical tunnelling and energy quantization in electrical circuits, since those are precisely the attributes that make a superconducting qubit work for computation.  

The work John is doing at Qolab, a superconducting qubit company seeking to build a million qubit device, is really impressive, as befits a Nobel Laureate and a pioneer in the field. In our conversation we explore the strategic shifts, collaborative efforts, and technological innovations that are pushing the boundaries of quantum computing closer to building scalable, million-qubit systems. 

Key Highlights

• Emerging from Stealth Mode & Million-Qubit System Paper:
  • Discussion on QoLab’s transition from stealth mode and their comprehensive paper on building scalable million-qubit systems.
  • Focus on a systematic approach covering the entire stack.
• Collaboration with Semiconductor Companies:
  • Unique business model emphasizing collaboration with semiconductor companies to leverage external expertise.
  • Comparison with bigger players like Google, who can fund the entire stack internally.
• Innovative Technological Approaches:
  • Integration of wafer-scale technology and advanced semiconductor manufacturing processes.
  • Emphasis on adjustable qubits and adjustable couplers for optimizing control and scalability.
• Scaling Challenges and Solutions:
  • Strategies for achieving scale, including using large dilution refrigerators and exploring optical communication for modular design.
  • Plans to address error correction and wiring challenges using brute force scaling and advanced materials.
• Future Vision and Speeding Up Development:
  • QoLab’s goal to significantly accelerate the timeline toward achieving a million-qubit system.
  • Insight into collaborations with HP Enterprises, NVIDIA, Quantum Machines, and others to combine expertise in hardware and software.
• Research Papers Mentioned in this Episode:
  • Position paper on building scalable million-qubit systems 

Creators and Guests

Host

Sebastian Hassinger

Business development #QuantumComputing @AWScloud Opinions mine, he/him.

Producer

Ayann Ettienne McGuire

Guest

John Martinis

John Martinis is a distinguished physicist renowned for his groundbreaking contributions to quantum computing. With a focus on superconducting qubits, Martinis has been at the forefront of developing high-fidelity qubits essential for scalable quantum processors.

What is The New Quantum Era - innovation in quantum computing, science and technology?

Your host, Sebastian Hassinger, interviews brilliant research scientists, software developers, engineers and others actively exploring the possibilities of our new quantum era. We will cover topics in quantum computing, networking and sensing, focusing on hardware, algorithms and general theory. The show aims for accessibility - Sebastian is not a physicist - and we'll try to provide context for the terminology and glimpses at the fascinating history of this new field as it evolves in real time.

Sebastian Hassinger: 00:30

Welcome to New Quantum Era, the podcast where I speak with people working in the field of quantum information science and technology. I'm your host, Sebastian Hassinger, and I actually have a first this week. This is a repeat of an episode that I recorded earlier this year, and this is triggered by the news last week that the Royal Swedish Academy of Sciences has decided to award the Nobel Prize in Physics to John Clarke, Michel Devoret, and John Martinis. They're sharing that award for the discovery of macroscopic quantum mechanical tunnelling and energy quantisation in electric circuit. And it so happens that I interviewed John Martinez for the podcast back in March at the APS Global Summit.

Sebastian Hassinger: 01:24

We had a great conversation. John is affiliated with the UC Santa Barbara and with QoLab, a startup that is building superconducting qubits in a very interesting industry alliance with other entities like Applied Materials, Quantum Machines, and others. And in March, they were just emerging from stealth, and John was doing a demo of some of the early calibration software of the Quantum Machines booth on the show floor, and we arranged to do an interview as well. I really enjoyed the interview. John has a a great way of explaining.

Sebastian Hassinger: 02:08

It's very accessible, very, very advanced work, and he is, I think, pretty unique in the field in that he blends a really strong sort of system arc architecture, system engineering kind of approach, along with very deep quantum physics and quantum science. So congratulations, John, for the Nobel Prize. Very well deserved and a great moment for the field of quantum information science and technology. And let's take a listen again to the conversation recorded back in March with John. When we talked last, you were still in stealth mode.

Sebastian Hassinger: 03:27

Right. And knowing that you were sort of out of stealth mode now, I was skimming back over that paper on, you know, building a scalable million qubit system.

John Martinis: 03:39

Oh, yeah. That we published a few months ago.

Sebastian Hassinger: 03:41

Right. Exactly. Exactly. Which is a great read. It's a really good, overview of the systematically sort of like what are all the

John Martinis: 03:50

go ahead. So did you get through the whole paper? It's a long paper. It's a

Sebastian Hassinger: 03:54

long paper. This morning, no, did not.

John Martinis: 03:56

Yeah, that's okay. But what I like is it has, it talks about the whole stack and whatever you're interested in, you can read that section and figure out what's going on. That's right. But the whole point of the paper was to really talk seriously about everything. Right.

John Martinis: 04:13

So that people can see that there's a plan. Obviously, change. Yes. Okay, as you do that. But, and also that we thought not just, you know, how to improve the qubits now, but how you go to a 10,000, a 100,000, a million.

John Martinis: 04:27

Okay. And what we're trying to do is build on the infrastructure and test it at a few qubits or whatever. Yeah. But it's gonna be upward compatible. Right.

John Martinis: 04:36

We're not gonna say, oh, we're gonna use integrated circuit fab. You know, you have to go much deeper into that to make sure that it's really gonna work. Yeah, yeah. So that was, you know, that's what we've been thinking for quite a few years now, as you said in stealth mode. And I didn't have the motivation to work on it until our collaborators worked on the whole stack and said, okay, it's really time to do that.

John Martinis: 05:04

But yeah, that was kind of one of the nicer papers I've been on in my career because it just pulls together all this thinking that we've been doing over the past decade.

Sebastian Hassinger: 05:15

And what immediately occurred to me was something that you said in our conversation a couple years ago that that really stuck with me in that, you know, you have a a systems approach Right. To the way you're thinking. You're not, you know Right. Experimental physics gets you so far, but to build something with this kind of long term roadmap, there needs

John Martinis: 05:38

to be an architectural sort of philosophy and systems design kind of approach. And that's really apparent in that paper. That's right. And thank you for noticing that because that's kind of what the big focus was. And again, you know, what is the long term vision?

John Martinis: 05:53

What do you have to build to do something? And then you back away. Now, I've been thinking about this for a long time, but the paper forced us to be real disciplined to communicate that properly. And I'm glad it got across to you because that was kind of our intent and to do that. And the other nice thing, of course, now we're out of stealth mode, and I'll talk about it in my talk later today, is we have a very unusual business model in that we're collaborating with a lot of semiconductor companies.

John Martinis: 06:28

So typically what the various quantum computing people like Google is they're big enough, they can fund the whole stack. As a startup, as a lever, we can't do that. So we collaborate with many people, and especially with many semiconductor companies who have really deep expertise in different areas. And that way you get to take advantage of that expertise. Specifically, they're going to be able to scale better because they know how to do.

John Martinis: 07:03

And they'll know how to keep costs down, which is like no one talks about costs for a quantum computer. And it's big, but the hope is we can manage that a little bit better. Now the hard part about this is you have to share and you have to collaborate and you have to communicate to this whole team you know the details of what you're going to do And that's kind of everyone kind of wants to hold their information tight and not share their infrastructure. But we decided that, you know, that's the best way

Sebastian Hassinger: 07:36

to move forward. Now I'm getting where the name collab originates.

John Martinis: 07:40

Yeah, that's really big was collaboration. Yes, you got it. That was kind of the ethos that we started with. Right. Is that we and this is not that crazy of a business because when you think about the computer industry, semiconductor industry, it was like IBM for many decades.

John Martinis: 08:03

And then in the 70s and especially in the 80s, instead of everyone doing everything, it all branched out and people did their different specialties. And there's a natural reason why that happened and it's better.

Sebastian Hassinger: 08:17

Well, I mean, reflects this sort of systems thinking right from the beginning. Right? You're you're anticipating this thing growing to a scale where one entity I mean, even in IBM, even if it at its peak still had a supply chain. Right?

John Martinis: 08:30

I mean

Sebastian Hassinger: 08:31

That's there still is this sort of large, multiparty kind of collaboration that needs to come together. That's right.

John Martinis: 08:38

Yeah. And I I would think that's the right model. Yeah. And I we think we're far enough along that it's time to switch. It's like the 70s in It's the semiconductor time to switch.

John Martinis: 08:52

And we're trying out that model, but it's actually been working pretty well.

Sebastian Hassinger: 08:58

That's great.

John Martinis: 08:58

We like it a lot. It makes sense. And then, you know, we, CoLab, access system integrator at the low level hardware end. Right. And then in our collaboration at the very high end, we have HP Enterprises.

John Martinis: 09:12

Right. Which makes supercomputers. Right. So they're going to be the system integrator at the high end. And when you think about their expertise in software and running complex workloads, that's really good collaboration.

John Martinis: 09:27

And we talk enough of the same language it it you know, we can do this.

Sebastian Hassinger: 09:31

Yeah. And it's also, I think, the last couple of years has been a dawning realization that to put these devices, these eventual future devices into productive kind of solutions, there's gonna have to be very tight integration with classical computing at scale. So so an HPE is a really good choice for that.

John Martinis: 09:52

Yeah. HPE is good. And the other thing that happened is in our paper that's coming out, hopefully in a journal, NVIDIA's now on the paper. And that would be a very good collaboration because, of course, they make powerful GPUs. And I think if you want to do the error correction decoding, let's say simply and powerfully, then that's the way to do it.

John Martinis: 10:17

Now they still have to show you can do it. It's not obvious. It's a hard problem. One of many. Yeah.

John Martinis: 10:23

And for example, like Riverlane, they're trying to build ASICs, which is maybe the right way to go, but it's certainly a good way to go. Right. We do NVIDIA. And then, you know, we'll see what's

Sebastian Hassinger: 10:35

I noticed also you're partnering with quantum machines because you you call out in that paper at multiple sort of inflection points how the wiring and the control systems are going to be one of the major stumbling blocks for for transitioning

John Martinis: 10:48

We this into kind of took it at every power of 10 from a 100 to a 10,000, 100,000. We talk about what you need to do and how think about that properly. And that was a very good way for us to kind of describe what's going on. So we have an integrated solution that scales to about 20,000 qubits. Wow.

John Martinis: 11:12

And then, you know, which is one module, and then you put together the modules tiling them together, which can then scale up, you know, to a couple 100,000 to maybe a million. And then, you know, that's a pretty big dilution refrigerator and maybe one

Sebastian Hassinger: 11:30

So that's all in one fridge you're expecting?

John Martinis: 11:32

Well, we think at least about 200,000 is one fridge.

Sebastian Hassinger: 11:37

So you definitely have some new approaches to wiring and controls. Yeah, one well, big coax per cubit is not going to count.

John Martinis: 11:46

Well, actually it's a brute force wiring. Really? But if you use the proper semiconductor, you have to use semiconductor manufacturing. So we're doing, you know, wafer scale integration and then we're using a flex for, you know, not coax with flex. Right.

John Martinis: 12:06

You know, when people know how to make printed circuit boards. Yeah. Okay? And then you use CMOS control electronic ASICs, which we need to investigate. And so I think if you just do the brute force scaling, you can imagine a path forward.

John Martinis: 12:23

Obviously, if we can combine things, and we have some ideas on that, where we can combine things at the right place, I think we can do that too. But that's a long time. But I would say the the the first thing we're doing is figuring out how to build a 20,000 qubit module with really good qubits.

Sebastian Hassinger: 12:44

Very cool. So so let that brings us to the qubits. You've obviously got a track record in working in superconducting qubits, transmons. What can you tell me about the way you're designing the qubits for CoLab?

John Martinis: 12:58

So this is a very complicated system engineering problem, and people have a lot of different qubits to do that. But getting the right control and the right crosstalk and whatever is difficult. So we think adjustable qubits, adjustable couplers is the way to go. Just this morning, I was listening to a very nice talk from a Chinese university who's been able to get this to work, too. Their numbers are good, and Google's done that.

John Martinis: 13:33

Other people have done that. There's really some good reasons to think that that's the way to go. So it's that, and we think we have a design, we've done the theory, where we can make the intrinsic errors, let's say the two qubit errors, that's the hardest thing, down in the 10 to the minus four range. If we can just get the coherence of like 200 microseconds, we think it could be in the three times 10 to minus four inches. And that's including all the strays and everything.

John Martinis: 14:08

So, of course, that's on paper. Paper qubits are the best.

Sebastian Hassinger: 14:11

They are the best.

John Martinis: 14:12

So we have to make physical qubits. We have a design in the refrigerator right now that we're testing. And various things look good. But we're starting the whole process just saying, okay, we have to make really good qubits, what is that design? What do we have to do?

John Martinis: 14:33

We understand that, we're going to focus on that. And then once we can get that to work, or a few qubits to work, then at the same time we're working on all the scaling so that hopefully we can do that relatively quickly. But there's a lot of control and software you have But to build to do that's the basic plan. We're not trying to build lots of qubits in the beginning. We're trying to build a small number but do it really well and build it with the future in mind.

Sebastian Hassinger: 15:07

So tunable coupling, coupler, tunable transmon, tunable transmon laid out in a lattice.

John Martinis: 15:15

Yeah, we think that the surface code architecture is good. If someone comes up with some better ideas, we'll think about it. But coupling to a lot of qubits is hard. Yes. And you take, for example, this is what I've been told by some people at Google, you know, I don't, haven't done the analysis myself, but they do an architecture where sometimes you connect to two qubits and sometimes you connect to three.

John Martinis: 15:45

Right. And, you know, that's really much easier to make something.

Sebastian Hassinger: 15:49

Yeah.

John Martinis: 15:50

But what I've been told is that if you have qubit dropouts, because we're not going to make a perfect chip, Right. Then that architecture is pretty intolerant to errors. Whereas the surface code, you can have, I think, one or 2% error dropouts, and then you can work around it because the connectivity is So this is an example of system engineering thinking that you can do things that is kind of simpler right now, but you really have to think long term, what's going to be more reliable. So reliability is a big And

Sebastian Hassinger: 16:27

when you look at over the last year, we've obviously had some really big stories around error correction and also sort of new iterations of superconducting qubit design. So not just cat qubits, but dual rail qubits and, you know, the, the Atlantic qubit quantum circuits, etcetera. Do you see sort of that low level cubit design as being something that's somewhat flexible over time like in other words your your grand plan is is the the whole system architecture but you could potentially swap

John Martinis: 16:59

out So a different our business plan is to get really good at fabrication. Okay. Okay, and you'll do that well and we're gonna work it out for our own Cubic design. However, we're going to generate the processes and PDKs and whatever so that if someone else wants to build their thing, then they can build it in ours. In fact, we're meeting someone today or tomorrow to discuss, you know, sure that we're going to come back.

Sebastian Hassinger: 17:34

And that collaborative set of partners. Yeah, it's a collaborative thing. Makes sense to have multiple potential qubit designs in that sort of

John Martinis: 17:42

large ecosystem. That's straightforward to do. And the way that we designed the business model and our ecosystem, then we can change that. So someone has a better qubit, then that's fine. Now, what I would say is, although there's all these nice papers out there, and this is great, I really like it, we need to explore, it's not clear to me that they're going to have the right system design to get everything to work right.

John Martinis: 18:10

You know, for example, with our design, we think we can get a good, high fidelity qubit gate with the intrinsic fidelity in the 10 to the minus six or so. So, you know, it's very, very good. With a thirty five nanosecond gate, a really fast gate, not quite as fast as the one qubit, but really And given we have a limited coherence time, and you always want to, you know, make a fast gate, then that's a very good design. Good design, four qubits, you can lay it out. So we're willing to talk to people, but we think we have pretty good design.

John Martinis: 18:47

But that's good. What you want to do is have a good baseline design, and if something better comes along, we'd be happy to pivot. And we'd be happy once we develop this, we'd happy to fabricate a chip for someone else so we can do that. We have no problem with that.

Sebastian Hassinger: 19:07

Out of curiosity, so you're saying 200,000 potentially in one fridge, and your your your your goal is sort of a million cubit system. Are you thinking about transduction at this stage?

John Martinis: 19:17

So so there's there's three ways to scale to, you know, millions. One is to have a big fridge. A really big fridge. A really big fridge. Okay.

John Martinis: 19:28

It's expensive, but yeah, it's a little bit ugly. It's not modular enough. Yeah. But, you know, okay. And then the couple of things has to do with transduction.

John Martinis: 19:39

Everyone's talking about optical. Right. And the problem is it's not quite there yet. Right. But people are working on it.

John Martinis: 19:46

And then we'd be happy if someone could develop that. Right. And the nice thing is building modules at 20,000 and then at, let's say, 200,000, it's very natural to think of those modules as the central CPU and the other modules as distillation for the T state, the A state generators, and then to network it in. So the surface code architecture allows you to build that modularity. But again, we have to make the optical communication.

John Martinis: 20:26

So people have gone a long way, but it's still It's one of the areas I watch with fascination. Even with Atom systems.

Sebastian Hassinger: 20:35

Yes, I know.

John Martinis: 20:36

I've learned a little bit about them, and I think people can do it, but the communication speeds are really I know it's

Sebastian Hassinger: 20:43

in hertz. Yeah,

John Martinis: 20:45

and then it's a lot of overhead and a lot of distillation you have to do. So, you know, that's a big research area. So we kind of have two different plans, and then we'll see. Now also in the paper, our theory people have said you might be able to do kind of error mitigation and break up the system classically in a clever way so that you might be able to tolerate a lower, smaller system. So, you know, we'd be happy to consider that too, and obviously we can build systems available to them.

John Martinis: 21:21

But, you know, right now we're at 100 cubits or so. And it's hard to to know what it's gonna be like, you know, in the future.

Sebastian Hassinger: 21:32

Yeah. I always think that about, you know, the the the about algorithms or applications. It's sort of like, well you have to sort of build it to

John Martinis: 21:41

see But what you do with can't build it to the application. I know. But we have to build.

Sebastian Hassinger: 21:45

It's a chicken and egg problem.

John Martinis: 21:47

Yeah, that's right. Standard problem we have to do that. So just to tell you a little bit about my talk. Yes, please. If you look at the rate of progress, let's say the Google group the number of cubes, it's a good proxy because they're making the cubes better.

John Martinis: 22:06

If you extrapolate that out as to when we'll get to a million cubits, okay, I will be dead.

Sebastian Hassinger: 22:15

I was guessing it was gonna be bad news.

John Martinis: 22:18

That's bad news, okay. What I also say is if you look at all the young people in the audience, they will be retired. Yeah. And so we're in this situation now where we really want to make it go faster. Right.

John Martinis: 22:33

And that's the nice thing about a startup company is you're able to think more creatively, out of the box. What do we need to do and try to assemble the team? You don't have that inertia. So yeah, we're basically for my personal goals. Is I want to

Sebastian Hassinger: 22:54

Self motivation is good job.

John Martinis: 22:56

Before I die. Okay. And I think we have a plan where we can rapidly accelerate it because of course, you know, it's nice that this is funded well, both from private government, private industry and the like, but this long horizon is not good economically. Right. And we're trying to figure out how to speed it up.

Sebastian Hassinger: 23:17

Capital has a limited attention.

John Martinis: 23:19

That's right. That's right. And originally, when we did the quantum supremacy five years ago, it was ten years. Well, we're halfway through ten years, and we got a factor of two. So it's a little slow.

John Martinis: 23:31

And I'm going say that it's generically hard. So our position paper was, you know, what's an approach to solve that problem? And it's also approach, like I said before, is how to make it cost effective. Right. And how to do it in a way that that makes sense that

Sebastian Hassinger: 23:48

I I was thinking as well, you know, mean, again, from the sort of systems architecture kind of point of view, I mean, it's it makes sense that what we're chasing is that moment when we can leverage everything we've learned about CMOS. Right? I mean, that's the the most successful human scale story that's ever existed. We we fabricate, like, trillions of transistors a day at this point. So, I mean, in in some sense, does it feel like superconducting qubits are are, you know, one entering in a race.

Sebastian Hassinger: 24:19

They sort of emerged around sort of February. But there's all of the qubits are kind of on that same race to try to get to the point where there's an inflection point where all of a sudden you can have that monolithic solution moment, when suddenly you unlock the secret sauce.

John Martinis: 24:34

So let me you're right in what you're saying, but it's a little bit more nuanced, but it's an actually interesting thing. You want to take advantage of CMOS, but if you go to a CMOS foundry, and we actually tried that in the last few years, and we got a piece of paper out. Right. But the problem is they have CMOS tools and they want to shoehorn your process into a CMOS process. Yes.

John Martinis: 25:04

Now for spin qubits, know, okay, you can do that, fine. But for superlang qubits, it's a little bit different of a process. So it's fine, but you still have to buy the tools and bring them up and all that. The way we figured this out, took us a while to figure this out profoundly, okay? We knew it intuitively from the beginning, is not to go to a foundry.

John Martinis: 25:30

And not, it's not a 300 millimeter foundry is the answer. Right. Got it. If you have you go to 300 millimeter because they have the latest most advanced tools. The latest processes are only in 300 millimeter and not in 200 millimeter.

John Martinis: 25:46

Right. And if you want to back port them, I see. It's too expensive. Right. Because, you know, they don't want to redesign, there's no market.

John Martinis: 25:53

Yeah. Okay. Yeah. The other thing is we have to make two new tools, new process. We don't get the process that exists, but you have to combine it in a tool.

John Martinis: 26:02

I see. So the secret is, is to go to a tool maker. I see. And in our case, that's Applied Materials. Got it.

John Martinis: 26:10

And that collaboration has been fantastic. Excellent. Okay. And it's basically because we can tell them what we understand about the physics. Got it.

John Martinis: 26:24

And they go away and say, yeah, we have these Versa tools. And then we work together. Right. And then we finally say, okay, these are the tool sets.

Sebastian Hassinger: 26:31

So that's upstream from the fab, the eventual fab of

John Martinis: 26:34

this. Yes, and the thing is, is they know how to do the various processes, and they have experience. And they all the time are selling to people who want to do something new and they have developed in labs. They're used to this. And what we're doing is within what they're normally doing.

John Martinis: 26:56

And the nice thing is the tools that we want to make this in are, let's say $50 to $100,000,000,000 I don't know what this is. And I'm going say that's even too expensive for Google or whatever until you're selling something, It's way too expensive. But you know, for them, they have these systems, they can retrofit it relatively cheaply and then try it out.

Sebastian Hassinger: 27:21

I see.

John Martinis: 27:22

And then they have a, you know, several billion dollar research clean room. Right. So for a modest amount of funding, working with the right people, and of course we have to convince them we knew what

Sebastian Hassinger: 27:34

we were doing,

John Martinis: 27:35

okay, fair enough, okay, then you could take advantage of this. So as a So it's

Sebastian Hassinger: 27:40

their tool, tooling R and D process is what you're plugging into essentially. Yes,

John Martinis: 27:46

that's right. And then they've been able, you know, if they have to bring it up, it's taken a while to bring it up, you know, six months. Right. But they can bring it up in a way that just blows your mind in terms of like the university fab you ever do, because they have real experts there. Yeah.

John Martinis: 28:02

And then also we're thinking about how to do certain things in a couple years, and they have experts on that too. So you can really put together anything. But basically every process step is way better and way more reliable. And the really interesting thing, and I'll be giving a talk on this later today, is we just make simple resonators, you know, okay, just to check out the materials. And that's actually took a while to get that to work right, okay?

John Martinis: 28:32

But the reproducibility of those resonators, like the resonant frequency, is way, way better than we ever did. And that's over a huge 300 millimeter wafer. And you know, if you go look at the specs of what they're doing, they have nanometer precision on various things and flat surfaces and big grains and all these crazy things that you can't even imagine doing. And that just comes out of their knowledge base.

Sebastian Hassinger: 29:00

It's spectrum from science done by postdocs and PhD students to engineering.

John Martinis: 29:07

They have decades and decades and of then we take advantage of it. So it takes a while to get there, but then in the end you have something that is really kind of amazing. Now of course we have to show that it's gonna translate to qubits. My guess is what's gonna happen is we're gonna translate to something that kind of people have already done on their, let's say their best qubits or best Right. But we'll be able to make it reliably.

John Martinis: 29:36

And we'll be able to make it over a big 300 millimeter wafer and get it with the parameters set right and get everything right. But there's a chance that be much, because it'll be more reliable, it'll also be much better. Right. So we, you know, we'll have to, we'll have to check that, but it looks really, really interesting.

Sebastian Hassinger: 29:55

That's exciting. So, can you tell me roughly when you'll have a system that you can start publishing results or getting people to kick the tires on?

John Martinis: 30:05

Well, okay, so first of all, from the working with in Taiwan, Academia, Syndicate, and Itri, an archive paper was posted yesterday on a process that gets rid of these liftoff junctions. And liftoff, of course, no one in their right mind of the semiconductor industry would ever do liftoff. It's very, very dirty. Maybe at the back end or something, but to depend on that's bad. So we publish that.

John Martinis: 30:34

We still have to make it better. So that's published. We have the position paper. And are, yeah, we're working on getting everything figured out. Now, one of the things, it's going to be better than publishing.

John Martinis: 30:50

Okay, we'll publish.

Sebastian Hassinger: 30:51

Good. Okay. But what's better

John Martinis: 30:53

than publishing is that our CoLab business model is to make available to us our good qubits and our good system. So the big thing we do is what's called the COLAB START system, where we deliver a system that's essentially equivalent to our best research grade systems. So we take all the knowledge of that and we reproduce it and we give it to you or sell it to you because We sell it and then people can, you know, use good cubes. Right. And it's part of our collaboration strategy because we think university groups, they can test things that maybe we don't have the time with, or they're going to test and maybe come up with new ideas.

John Martinis: 31:42

That's great. And then besides the qubit systems, we are going to make available chip mounts and some filters and some electronic testing so that people can make sure that their systems are kind of up the grade. So there's just a lot of infrastructure to make good qubits and we'll try to make that available.

Sebastian Hassinger: 32:08

And when does that start? Start.

John Martinis: 32:11

We're hoping actually last night we talked to a group and they said they wanted to buy a chip mount from us, hopefully in the next few weeks. And the collab said there's another group that's talking about it and we hope to get them out of quote and get a thing in the next month or so.

Sebastian Hassinger: 32:30

Great. So this is going be a big year for you then?

John Martinis: 32:32

Well, hopefully if we can get all this to work. Excellent. So that's good. And then as we make the chips and we've tested it enough for ourselves and maybe, you know, publish some papers, it's going to be better because it's like the super appendix where you actually get a chip. Right.

John Martinis: 32:53

Okay. You know, that can play with yourself. Right. Right. And, and then I would say one of the things we're trying to do is to make these qubit qubit, good qubit systems requires a lot of patience and calibration and the like.

John Martinis: 33:09

And we're working closely with quantum machines to do that right, make sure the control systems are good and doing that. But we're trying to fix various things in the CUBA chips at the low level so that when you go to calibrate it, it's going to be a lot easier. Because what we did at UCSB and at Google really required a race car driver to drive the chip. And we want it to be, you know, not quite a Tesla self driving, full self driving, but you know, towards that, we want to move in that direction.

Sebastian Hassinger: 33:48

I mean, Tesla's just moving towards that as well.

John Martinis: 33:52

Fairness. Fine, fine. You understand, we want to move towards that. And it's not just making better software, want to make better hardware. So, you know, as an example, we're collaborating with Synopsys who makes EM simulation And design we're working with them to figure out how to do an EM simulation on, you know, a thousand qubit device or something.

John Martinis: 34:23

It's actually a hard problem because of the different length scales involved. Right. And, but they have, some knowledge base and they've done it on other kind of problems and they just have to apply

Sebastian Hassinger: 34:35

it to That's fantastic John. I think it sounds very much like you found a model that is going to have sort of multiplying positive effects on whole ecosystem because everybody is involved in that partnership.

John Martinis: 34:48

Yeah and okay, we're going to help other people make money and that's fine, okay. I'm just excited. We want to have a valuable company. We think we'll do that. We think that's the way to do it.

John Martinis: 35:01

And I think we have to do this because I think the physicists are being a little bit naive of what it takes. I don't know exactly what everyone's doing. I think a horizontally integrated, a lot of collaboration is going to help

Sebastian Hassinger: 35:17

the That's field a fantastic. Well, we'll be watching.

John Martinis: 35:19

Oh, and by the way, it's really fun. Of course. And because I get to talk people who are experts in their domain. Yeah. And then we get to learn about all these things in great detail.

John Martinis: 35:32

Right. You know, in a way that you could never hire someone in your own company. Of course. With that kind of expertise. Yeah.

John Martinis: 35:39

And then also I find is talking to the managers of the other companies, they're really good and they're used to managing a hardware project, which is different than managing a software project. Right. And that's, you know, at least at our level, that's fine. And it's really interesting to learn from them. So, so yeah, I get to get to experience things, which I find really, really fun.

John Martinis: 36:05

Matches with, I mean, you, you tend to have

Sebastian Hassinger: 36:08

a very holistic perspective and that's the approach you found with CoLab. It sounds Yeah.

John Martinis: 36:13

That's right. A terrific model.

Sebastian Hassinger: 36:14

Thank you so much for joining us. This has

John Martinis: 36:15

been great. Excellent. Great. That was fun.

Sebastian Hassinger: 36:36

By OCH. You can find past episodes on www.newquantumera.com or on blue sky at newquantumera.com. If you enjoy the podcast, please subscribe and tell your quantum