Go ahead and get started. Welcome everybody to the second day of Needham's twenty-first Annual Technology, Media, & Consumer Conference. My name is Quinn Bolton. I am the quantum computing and semiconductor analyst for Needham. It is my pleasure to host this fireside chat with Rigetti Computing. Joining me on stage is Dr. Subodh Kulkarni, President and CEO. We also have CFO Jeffrey Bertelsen in the audience in case there are any tough financial questions. I know, you know, quantum computing companies have been public now for three or four years, but there may be some in the audience that are still less familiar. So maybe, Subodh, just to start us off, can you just give us a brief history of Rigetti Computing and why you chose the superconducting modality for your qubits?
Sure. Good to be here, Quinn. Rigetti started in 2013, as started by Chad Rigetti, who did his PhD at Yale University in superconducting quantum computing, then was with IBM Research for a few years and then went off on his own and started Rigetti. Went through the typical startup. He based it in Berkeley, California, which is where we are based. Went through the typical startup journey of seed money and in series A, B, C. Series D is when the company decided to go public through the de-SPAC process. The company went public in March of 2022, and then we have been public since then.
The main reason, Chad and we started with superconducting, and we continue to believe in superconducting, particularly superconducting gate-based modality, is because of the benefits of superconducting modality in terms of scalability and speed. We are dealing with semiconductor chip technology, we know it's relatively easy to scale up. Because we are dealing with electrons, we are dealing with gate speeds commensurate with CPU and GPU gate speeds, we did talk about 10s of nanoseconds. Those are the main reasons why most of us in the superconducting gate camp continue to invest in superconducting. Our challenge is fidelity. That's where because we are dealing with man-made chips, that means there are errors, and that's where we usually struggle when compared to other pure modalities.
One of the questions we get is, Rigetti and the superconducting camp competes against companies like Google and IBM that are obviously much larger with greater financial resources. How do you compete against those companies with greater resources? You know, how do you stay ahead?
Certainly the fact that we started in 2013 and we have a lot of good patents that have already been issued is a key differentiator for us. We are roughly at about 300 patents right now. There are many areas of quantum computing that Rigetti pioneered. More recently, the specific area where we differentiate ourselves from IBM, Google, are open modular architecture. IBM and Google have chosen to build a quantum computer like a mainframe computer. If you look at their system, they do everything, the chip design, the chip fabrication, the various layers of the hardware, the various layers of the software, including deployment on the cloud. Everything is controlled by IBM or Google. We have taken the exact opposite stand. Ours is open modular, we can integrate third-party solutions relatively quickly.
That we think it's a right thing to do because it allows the ecosystem to grow. It allows innovative solutions from outside to be integrated faster. For instance, we are on the cloud deployment side. We are partnered with AWS and Microsoft Azure. I mean, all fairness and credit to IBM is not exactly a cloud player. AWS still has 40-45% market share. When it comes to the distribution layer software, we partner with Nvidia because like it or not, CUDA has 95% plus market share of data center distribution layer software today, whereas IBM and Google are going to compete with Nvidia in that area. We think it's actually better to accept where there are well-established giants in the stack and use them rather than try to do everything yourself.
That's one big area of philosophical difference between us and IBM, Google. It gets into the more details and how we design our chip and how we build a physical QPU itself. Rigetti pioneered the concept of chiplets in quantum computing, and that's a key differentiation for us. Fundamentally, we use chiplets in semiconductor industry all over the place because it's easier to build smaller chiplets than single large chip. We found the same thing in quantum computing. Several years ago, we were the first ones to jump onto the concept. We have several key patents in that area that prevent others from just jumping into chiplet architecture. Right now, as far as we can tell, we are the only company who has physical quantum computers deployed on the cloud or in on the deliver with chiplet architecture.
We fundamentally remain convinced that's the right way to scale up. Right now we are all at about We just deployed our 108 qubit system a month ago or so. IBM said 120 qubit. Google said 105 qubit. Ours is a chiplet-based architecture, we have 12 9 qubit chiplets. IBM and Google have a monolithic chip. We think it's going to be extremely difficult for IBM and Google to get to a 1,000 qubit with a monolithic chip, whereas we believe the path we have to get to 1,000 qubits is not easy by any means, but easier than a single chip approach. There are other areas of differentiation too that get into the gate structures and annealing processes and so on, I won't go into all the details.
Fundamentally, I would say our open modular architecture and our chiplet architecture are probably how we differentiate ourselves from companies like IBM or Google.
Yeah. That Chiplet versus monolithic certainly seems like a key differentiator, and I know in semis it's a lot harder to yield a big chip than.
Exactly
multiple smaller chips, it seems like the same is true in quantum. Just looking at the industry and investment in quantum, I think we're seeing growing support from governments around the world and growing adoption or at least interest in enterprises. Maybe just starting off, what are you seeing in terms of overall interest in quantum? Then any thoughts on the U.S. government, will they ever reauthorize the NQI Act at higher levels of funding? It feels like we've been in Congress, you know, for a couple of years now waiting for that reauthorization. Sorry for such a long question, but maybe also talk on the U.K. government's ProQure initiative.
which sounds like it's got a fair amount of funding set aside to invest in quantum.
Yeah. I mean, overall, if you look at the interest in quantum computing, it continues to grow very rapidly across the world. Driven primarily by governments and national labs, but now commercial entities are getting into quantum computing too because they see the value of quantum computing. Overall, the interest is soaring quite rapidly. Without a doubt, the biggest investment from national lab or government standpoint seems to be happening in China. The government of China, they are being very secretive in this area, but all indications are numbers in the neighborhood of $2.5, maybe even $3 billion per year are being invested by the government of China in quantum computing. By the way, most of that is in superconducting quantum computing from what we can see. Maybe a little in photonics, but mostly in superconducting quantum computing.
That seems to be the case. Once you leave China aside, without doubt U.S. government is a big player. There are many other countries that have their own quantum initiatives. Speaking of U.S. government itself, you correctly pointed out the original NQI Act that was passed in 2018, expired in 2023. It's been a couple of years, more than a couple of years now. We are eagerly waiting for NQI reauthorization to pass. It's still expected to happen sometime soon.
Right. Right.
I hate to say it again. We definitely expect some version of NQI reauthorization to pass. In the meantime, there has been some line items that have passed with some reconciliation bills. DOE has freed up about $400 million right now, so that's already appropriated, and some projects are coming out of that line item. DOD continues to increase the individual line item activities in the defense bill, so we are definitely benefiting from that through contracts like Air Force Research Lab and so on. It's not like the U.S. government is spending zero by any means.
Right.
It's just not as well coordinated as an NQI reauthorization bill would pass. The expectation is that once the NQI reauthorization bill passes, U.S. government will be investing roughly $1 billion a year in quantum computing. Right now the number is probably a third of that, but we definitely expect that to happen. You go around and say, U.K. government, a lot more streamlined. You're right, the U.K. government has a ProQure initiative. It's a 6-year program that they have very clearly laid out the milestones over the next 6 years, and it's starting now. They have passed the bill through their parliament and funded it. It's somewhere in the neighborhood of about $300 million-$400 million a year. The overall number is less than the U.S., but it's more focused and streamlined.
You go around, we recently got an order from the country of India. Indian government is doing something similar. The number is smaller, about $250 million or so per year. It all adds up. You go around and you look at individual European countries, Germany, France, Italy, Spain, even the Scandinavian countries, certainly Japan, Korea, Australia. Almost all the developed countries in the world have a quantum initiative right now for obvious reasons, because of national security interest and so on. The numbers vary, but when you add up, the numbers are becoming sizable. We are talking about, even keeping China out, we are talking about $2 billion-$3 billion a year type numbers just driven by government and national labs.
It's exciting and the number is bound to grow as we get closer and closer to quantum advantage and commercial businesses start coming in too.
Right. We'll get into quantum advantage and timing on that. I imagine that as you get closer and past quantum advantage, the commercial or the enterprise business.
Absolutely. Thank you
really starts to ramp. The company reported earnings on Monday night. Since we're just coming out of earnings, maybe just spend a minute talking about what you thought some of the highlights were for the 1st quarter, some of the achievements that you announced on the call.
The main takeaways we focused investors on during the earnings call were 3 things. First was, we successfully deployed our 108-qubit system on cloud, including AWS, Microsoft Azure, and others. That's a huge accomplishment. I mean, it's one of the most powerful quantum computers in the world right now. Only second to IBM's 120-qubit system, ahead of Google's and well ahead of any other modalities or so on. We are very proud that system is already available for anyone to use and we can already see the increasing interest at AWS and Azure and other platforms, that's very exciting to see. That was 1 big first takeaway.
Second one, we touched on it, we are seeing increasing traction with government national labs, universities, but also commercial outfits. You see that in sales. I mean, we didn't make a big deal of it, but our sales did triple year-over-year. We definitely are seeing more and more traction from government national labs and commercial organizations, so that's good to see. Last but not least, technology roadmap. We wanted investors to continue to focus on technology roadmap and the milestones. We are roughly at a 108-qubit level, roughly at about 99.9% one-qubit gate fidelity, 99.1% two-qubit gate fidelity, 60 nanosecond gate speed. As good as it sounds, it's not good enough to deliver practical quantum advantage for practical applications.
We think we have roughly a 3-year timeline to get to the 1,000-qubit, 99.9% 2-qubit gate fidelity. It's important we focus on those technology milestones because that's really what's going to enable quantum computing to take off. That was the third takeaway. That's what we focused our investors on.
How should investors be thinking about those technology milestones, say, in 2026 with Cepheus 108? How much further do you think you can increase the fidelity target this year? I know ultimately you wanna get to 99.9% in 3 years, and then I think there was also a discussion previously of a 150 or greater than 150 qubit system potentially in 2026. Is that still on the roadmap?
You're right. If you look at where we are today, 108 qubits at, if you just focus on the 2-qubit gate fidelity, 99.1%, and gate speed about 60 nanosecond. We know in roughly 3 years we want to get to a 1,000-qubit, 99.9%, and less than 40 nanosecond gate speed. The easiest one is gate speed between the three. We feel pretty good about that one, to get from 60 nanosecond below 40 nanoseconds. After that is the qubit count. It's not trivial, but with chiplets, we feel pretty good about getting to 1,000 qubits in roughly 3 years. The first is the fidelity for us. We are at 99.1% 2-qubit gate fidelity. We need to get to 99.9% in 3 years.
Development is not a linear process, so there will be some staircasing going on. We need to make some concrete improvements in our technology to get to the next year. Right now, one of the key areas that is limiting our fidelity is coherence time. Coherence time is the amount of time a qubit stays in its quantum state before it loses its quantum state. That's the time we have to do the computing action. Obviously, longer is better. Right now, our coherence times are in the 25 to 30 microsecond range. That's how long the qubit is staying. We need to improve that to by a factor of 2, ideally a factor of 3, to reach 99.5%. Then beyond that, there will be the next things.
When you're doing these kinds of developments, you don't, you never completely get anything done.
You're never totally finished.
Yeah
it sounds like.
Even in semiconductor industry, after 60 years.
Yeah
we, if you look at what semiconductor industry does, it shrinks the transistor and improves the fidelity, and that's what happens year after year after year. This is going to be an ongoing conversation. It's not going to be like, "Oh, we are done with coherence time." Every year we will have to improve the coherence time a little bit, improve the fidelity a little bit. While we increase the qubit counts, there will be other problems that come in. This is going to be an ongoing development process. Anyway, right now our focus is to improve coherence time, get to 99.5% by the end of this year with the 180-qubit system while we continue to increase the qubit count. Definitely we are working right now on 150, 700 qubits.
Frankly, in our Fremont site, we have started purchasing dilution refrigerators that can handle up to 1,000 qubits. We are even experimenting with some 1,000 qubit type approaches right now. As we get closer and think that the systems are deployable, we'll obviously be announcing those systems. Definitely expect us to increase both qubit count and fidelity as the year goes on.
Okay. One of the things I think highlighted on earnings over the last quarter or two is an increasing number of system deliveries, maybe just talk to us about the roadmap for system deliveries this year between Novera QPUs, Novera systems, as well as the C-DAC 108-qubit system later this year.
Yeah, getting into this year, we told everyone that there were basically 3 big systems we had to deliver for on-premise. 1 was to an Asian company, 1 was to a Bay Area company, and the third was to C-DAC in India, the government national lab. That was a 108 qubit system. The other 2 were 9 qubit systems. Those were the 3 on-premise orders. Along with it, we continued to talk about Novera 9 qubit system. We announced that we got a 9 qubit system order from a Japanese research organization. We got another order from University of Saskatchewan in Canada. We'll get a few more orders like that as the year goes on, and we'll disclose them at the right time. If they give us permission, we will give their names.
Sometimes they do, sometimes they don't, as you can see. Certainly that is helping our growth, and we'll continue to do that. Our, our focus primarily is on the technology milestones while allowing customers to experiment with our products. We know, maybe we choose our customers too. We want good quality feedback to come back to us, so we don't want to just give a system to somebody and then they disappear. That frankly doesn't help us in the stage of R&D we are in. We want customers like University of Saskatchewan. They are creating a Canadian quantum ecosystem around it, including some specific technologies that the government of Canada is interested in investing in, such as dilution refrigeration technology.
There's a very innovative company in Canada called Zero Point Cryogenics that have very clever concepts in dilution refrigerator, and we want to learn from those concepts. We want to see what exactly the benefit of that dilution refrigeration technology is compared to what we use normally, which is either a company from Finland called Bluefors or a company in the U.S. called Maybell Quantum. We want to learn as we continue to refine our quantum chip itself. We will continue to do that. We'll continue to talk to key research outfits to help us get high-quality feedback.
That feedback loop.
is very important at this point.
is very important. You had discussed on the call, but I think originally announced back in March that you intend to invest $100 million into the U.K. over the next several years. I think some of that is to position yourself for the ProQure program. What are you spending or what will the $100 million investment entail in terms of people and system deployments in the U.K.?
Yeah. Indeed, that's an exciting announcement from us, consistent with the U.K. government's ProQure initiative. We have said we'll invest about $100 million over the next 3-4 years in the country of U.K., and that's an all-encompassing. It includes CapEx, physical quantum computers we will install in the U.K. People, we already have a small group of employees in U.K. We'll increase that head count as well as facilities. Right now, if you go to U.K., they have the big quantum computing development center over there is National Quantum Computing Centre, NQCC, which is located outside Oxford. If you physically walk It's a beautiful building that they built for $ several hundred million about 2-3 years ago. The highlight of that building is actually our quantum computers.
If you walk in there, you will see a Rigetti 36 qubit quantum computing system, being used by hundreds of researchers at NQCC and across the entire U.K. ecosystem, Oxford, Cambridge, and other universities. They have some fantastic quality R&D going on. Let's not forget that the key transistor, if you will, that we all use in superconducting quantum computing is called Josephson junction, and Josephson was a physicist at University of Cambridge.
Okay.
Even to date, he is in the record books as the youngest physicist to ever get a Nobel Prize. He got the Nobel Prize at the age of 27.
Wow.
Okay? He's still there at University of Cambridge. Let's not forget U.K.'s contribution to quantum computing. They are one of the key contributors to quantum computing right now. Anyway, we are very proud to be the flagship of NQCC's quantum ecosystem right now. The ProQure program is starting now. It's a very systematic program. It's a six-year program. The first phase of companies they will choose will get roughly $90 million over the next two years. The next phase, the companies will get about $100 million over the next two years, and so on. Implicit in our statement of investing, we are also assuming that we get the benefit from ProQure initiatives. We definitely are optimistic that we will be in the first phase and the second phase, and so on.
Our announcement is consistent with the ProQure program. As we expect to get awarded that contract, we are committing that we will bring quantum computers in the country, just like we already have, but increase the staff, rent out facility, and stuff like that.
Okay. The timing for the ProQure, the first phase, I think you said on the call was maybe over the summer or July, August timeframe?
Yeah. The timeline right now is July, August time period is when they will announce the first phase of companies that they go with. We are optimistic we are one of them, but obviously there's stiff competition. There are 100 plus companies in quantum computing, and I'm sure many of them will be applying for those grants and awards. We'll see how it evolves.
Okay. Well, good luck. We will stay tuned. You talked on, you know, here already just the correlation between coherence time and fidelity and your efforts to try to double, if not triple coherence time over the course of 2026. Without maybe getting too technical, what types of things are you doing to increase coherence time? Is it more on the manufacturing side? Is it on the chip design? How do you improve the coherence time?
Coherence time is controlled by a lot of things. It's a system variable, if you will. It's controlled by a lot of things. Our primary focus right now is on the chip design and fabrication. When we design the chip, there are many things we do in coupling the chip. That's how we put things in a quantum state. The precise definition of the pad, the lines that are driving the qubit, so the areas and so on. We are changing a lot of those things to improve the coupling of the chip, if you will, with the lines, and that has a big impact on coherence time. Also the fabrication processes and the materials too.
One of the areas that we have actually disclosed, and this is joint IP that we actually have with Fermilab, we use niobium contacts right now for superconducting contacts. Niobium tends to oxidize fairly quickly. We did some work with Fermilab to show that if you cap niobium with tantalum tends to oxidize a lot less. You can see about a 50% improvement in coherence time. Now, that was done in a fundamental way. Now we are incorporating that in the actual chip itself. That would be one example. There are other process steps that we are doing. Right now, if you look at the cross-section TEMs of our devices or other devices too, you see intrinsic roughnesses in that Josephson junction area, which is the critical part of the quantum chip.
We need to improve the roughnesses and the point defects, there are things we are doing in the deposition process, oxidation process, etching processes to smoothen out those surfaces. All those things add up. Individually, we have seen that one of those things can add like 50% or 70% improvement coherence time. It's different to do that in a one-off TIM mode versus in a final finished chip mode. That's the work we are doing right now.
Okay.
We are pretty sure we'll improve the coherence time as the year goes on, and that will definitely reflect in fidelity numbers.
it sounds like a lot of it is sort of the semiconductor processing.
Processing
processing and improving the steps or film uniformity or not uniformity, but smoothness and steps like that.
I feel like I've come a full circle if I go back.
Yeah
I mean, I started my career in I was working at IBM CMOS, and my first job when I finished my PhD and joined IBM was to improve the smoothness of the CMOS oxide, and to improve the point defects and to improve the yields that we were getting with CMOS. I feel like I've come a full circle now working on similar things but with quantum chips now.
Yeah.
It's essentially the same physics that we are dealing with. Anytime you have point defect or some kind of a grain boundary defect, it can lead to catastrophic losses, and we are dealing with similar issues. That's what we need to improve on.
Okay. It sounds like fidelity is probably the greatest challenge or area of greatest focus to hit your milestone of the 1,000 qubit system, 99.9 qubit. Are there any other sort of major challenges you see getting to that quantum advantage system in about three years?
Getting coherence time, definitely we talked about it, has is one of the top ones. Chiplet allows us to scale up relatively easily, but it's not trivial. I don't want to say it's just a slam dunk by any means to get to 1,000. In addition to chiplets, Right now in our 108-qubit design, we use a monolithic cap. The cap wafer itself is where the connections are coming from, and each chiplet, the multiple chiplets get bound by a single monolithic cap today. Because of the physical limitation of the packaging equipment we have, we need to go to multiple caps for 1,000-qubit.
It's not as difficult as going from a single chip to chiplet, but there are some tricky parts involved when you are going from a monolithic chip to a multi-chip design. That's the next big hurdle we have to overcome in the next year or two, is to go from a single chip to a multiple chip. And should be easier than chiplet, but it's still got its share of challenges. I would say improving coherence time, top one. Going to multiple chips, the next big one. After that, who knows what else we encounter as we get the actual data. Development is never quite clear until you get the data and see what the numbers look like.
Maybe I'll skip forward because I had a question about the fab and the investments in the fab. As we talk about improving coherence times, we talk about the multi-cap, it does sound like a lot of this is manufacturing related. What do you think the requirements will be on the CapEx side for fab 1? Do you have the equipment to get to 1,000 qubit systems? Are there investments needed? You've talked about the possibility of a second fab.
At some point in the future, maybe just any updated thoughts on manufacturing?
Sure. Right now we have a full-fledged fab in Fremont, California, that we use on a daily basis, and we have invested more than $200 million in it over the last five, six years, and it's doing a great job in giving us the chips and everything we need, including packaging equipment. We will need to continue to upgrade it, those numbers are not significant. We are talking $10 million, $20 million, $25 million a year to keep a fab upgraded reasonably well. The bigger investments that are going right now are in the 1,000 qubit systems. There we need bigger dilution refrigerators. A single 1,000 qubit system, I mean, each qubit has multiple lines going to it. These are typically coax lines or even flex lines.
The signals are extremely weak, we need to amplify the signals. We need these special amplifiers that we call traveling wave parametric amplifiers. These are not off-the-shelf amplifiers we can buy. All that stuff is extremely expensive. When we look at 1,000-qubit system, even keeping the chips aside, the BOM, if you will, for a company like us, we are talking $20 million-$25 million just for cables, amplifiers, refrigeration systems, and so on. To do the work next year with 1,000 qubits, we are buying multiple of those systems. You cannot just have one of those systems. Definitely expect CapEx to increase this year because we are investing in 1,000-qubit type systems and all the associated infrastructure around it, along with the fab. We are not taking huge jumps.
I mean, the, our overall burn rate, if you will, in 2025 was about $75 million-$80 million. Expect that to go up by 15%-20%, maybe a little more, depending on the exact timing of all these orders and everything. We are not taking a factor of two or anything like that.
Right.
Considering our cash position, we are at about $570 million cash at the end of Q1, no debt. We still feel very good about our longevity of cash, including the CapEx, additional CapEx we are incurring right now.
Got it. It sounds like a lot of that CapEx is actually almost more outside of the fab for other systems, dilution refrigerators.
It's mostly for 1,000 qubit systems.
control systems, lines.
Yeah.
Yeah, okay. You touched on one of the strengths of the superconducting modalities, the gate speeds, and I think the Cepheus 108Q is about a 60-nanosecond gate speed.
You've demonstrated gate speeds as fast as 28.
28, yeah.
How does that compare first to the gate speeds of Google and IBM? Maybe for folks who may not be as familiar, what are the gate speeds for some of the other modalities?
Sure
you know, the neutral atom or the ion trap?
As I mentioned earlier, one of the main advantages of superconducting, which is why you look around the world, and I would say more than 90% of the investment in quantum computing is going in superconducting modality right now. It's, look at government of China and other countries and everything. The main reason is speed and scalability. We deal with, right now our 108-qubit system is at 60-nanosecond gate speed. If you compare it to trapped ion or pure atom, they are talking about 500, 600 microseconds. Almost 10,000 times slower than superconducting, which for them is a huge challenge. It's simply physics. We are moving electrons, they are moving ions or atoms. By definition, ion or atoms are thousands of times bigger than electron, it's just physics dominating their speed.
They have a tough time getting over that barrier. Whereas we intrinsically, electrons, which is what CPUs and GPUs use, we always will have that strength. That's why we like the modality itself. Now, within the superconducting camp, IBM, Google, and us, we had slightly different philosophies a couple years ago. Google and we always believed in what we call a tunable coupler technology. The coupler, you can adjust the frequency of the coupler to get the most optimum interactions between the qubits, and that allows us to get faster speeds. For a long time, IBM stayed with what they call fixed coupler technology. It's easier to build a fixed coupler chip. You can get more qubits because a tunable coupler chip is a lot more complicated.
You see that a little bit in IBM's qubit count was always higher than Google's or ours because their chip was relatively easy to build compared to ours with the tunable coupler. Recently, IBM also moved to tunable coupler technology because the strengths in tunable coupler are better than fixed coupler. You see that IBM actually went down in qubit count. With their fixed coupler, they used to be at 156 qubit, now they have come to 120 qubit. You see the impact. The benefit is IBM's gate speeds used to be like 150 nanoseconds with fixed coupler. Now they are in the same league range as where we and Google are, in the 50, 60 nanosecond range. Really it's the physics of the tunable coupler that is dominating our gate speed right now.
All of us, Google, IBM, who are using tunable coupler technology, we are in the same ballpark. In the meantime, we announced that we have come up with a new gate scheme that we call adiabatic CZ gate scheme. It gets very technical, but it's a very clever way to communicate information between the qubits, and that allowed us to get another factor of 2 improvement in gate speed. We demonstrated 28 nanosecond gate speed with the adiabatic CZ, and that's what we are starting to work on right now. As the year goes on, definitely expect us to deploy that gate scheme. There's a lot of innovation going on in universities. I'm sure IBM and Google are working on different kinds of gate schemes too.
We definitely expect them to be announcing something similar, maybe different in terms of the exact implementation, but we'll continue to innovate in that area. Gate speed is a key advantage of superconducting systems, and we'll continue to exploit that.
Yeah, continue to push that. Okay. One of the questions we've gotten with IonQ's announced acquisition of SkyWater, I think there's some concern that SkyWater, which was sort of served as a quantum foundry for a number of folks in the industry, that as that capacity goes to IonQ, that folks may be looking for other sources of foundry capacity. I know you have a fab, you've done some foundry work for government labs and agencies. What are your thoughts on potentially doing more commercial foundry work to the extent that capacity at SkyWater is, you know, affected by the acquisition?
Yeah. Our view is our current fab is mostly for internal use. We do offer it as a foundry service, as you pointed out, to select customers, primarily the U.S. government and the U.K. government. We serve as a foundry for organizations like Fermilab or Air Force Research Laboratory or to National Quantum Computing Centre in the U.K. Going forward, we will do that similar thing with the Indian government now with the C-DAC orders and so on. We don't want to take commercial customers because that just hurts us in our capacity. In fact, even with the government customers, we have a restriction on how many wafers per month they can run.
Okay
and they're usually respectful of that restriction. Could we offer it to more customers? Potentially, but it gets into what exactly do we get benefit. I mean, we don't want to make the small amount of dollars for that kind of service. There's a lot of IP and knowhow involved in the fab itself. We want to be careful to offer it as a foundry service. Frankly, look at the big companies right now. IBM has their own fab. Google has their own fab. Amazon just built their own fab. Most of us who have been in this business understand the importance of a fab.
Right.
We, all of us built our fab has been around for more than five years now, along with IBM, Google, and others. Most of us understood the importance of having your own fab because it's such a critical component of-
Because you can spend-
time to market. Exactly.
wafers and, yeah.
Exactly.
You'd mentioned the open, modular approach of Rigetti systems versus the closed systems at IBM and Google. One of your partners that you've announced a collaboration with an investment from is Quanta Computer, they're working on control systems.
Talk about how Quanta has, you know, progressing in its control systems? I believe you're using some of their control systems. in your Novera systems. Ultimately, when do you think they'll be capable of, you know, control systems for higher qubit machines?
Sure. We announced a strategic partnership with Quanta little more than a year ago now. They invested some money, about $40 million in Rigetti, but more importantly, the long-term commitment of, like, $250 million of R&D in the hardware stack, specifically control system. They've done their part. We obviously educated them on how to build a control system for our kind of modality. Quanta is a phenomenal company. I mean, they are a huge company. They know how to build CPU, GPU servers for cloud. They're Nvidia's closest partner. They have done exactly what we expect them to do, which is bring in volume manufacturing understanding. They have improved the performance of what we were doing, which is no surprise given their capabilities. But more importantly, they have slashed the cost dramatically.
When we were building control systems in Berkeley, you can imagine what the cost was with PhD-level engineering talent. Quanta has brought their classic high-volume manufacturing expertise in Taiwan. They have improved the performance and dramatically lowered the cost of control system, which is good on both fronts. We are already starting to use their control systems for 9-qubit type systems. They're not quite capable to do 36-qubit or 108-qubit yet. That's what the goal for this year is to Quanta to get up to that. Once they are completely up to speed, we'll start dialing down our effort in the control system area and focus more on the chip design and fabrication.
You save on R&D.
Exactly
you save on BOM cost.
Exactly
pretty good-
It's a good deal all around.
good relationship.
Clearly Quanta is going to be one of our close manufacturing partners as volumes increase, and I would say none better than Quanta in that world. They are one of the top ODMs in the world right now.
My last question is, we're getting close to the time, in April, NVIDIA announced its Ising, calibration- model that, you know, can help replace manual tuning of qubits. As your systems get to 100 and approach 1,000, is automated calibration something that could benefit you, either in terms of fidelity or just reducing the bring up time of these systems? You know, is that something that's of interest?
Yeah, absolutely. The NVIDIA announcement was important on both counts. They announced the Ising open-source models for calibration and tuning of a quantum computer, but also on error corrections. Two different things, right? Both are important for us. Definitely the calibration tuning, we are looking at how they can speed up calibration and tuning using NVIDIA GPUs. Equally important is, and probably more important, is the use of GPUs for error correction. Right now, what we do, along with our partners like Riverlane, is mostly FPGA-based error correction. If you look at our entire architecture, it's FPGA based. NVIDIA has come up with some very interesting concepts of using GPUs, should be no surprise.
Yeah
on how GPUs can accelerate that, and they claim a factor of 2 to 3 benefit with that. All of us are working with them, trying to make sure that indeed we see a benefit of 2 to 3. If we do, we absolutely will start using it. That work is going on right now.
Excellent. Well, we are at the end of the time for the session. Subodh Kulkarni, thank you very much for joining us at the AIME Conference. We really appreciate your participation.
Thank you, Quinn Bolton.
Thank you. Thank you, everybody.