管理層發言
Good day, and thank you for standing by. Welcome to the Rigetti Computing Third Quarter 2025 Financial Results Conference Call. Please be advised that today's conference is being recorded. I would now like to hand the conference over to your first speaker today, Dr. Subodh Kulkarni, Chief Executive Officer. Please go ahead.
Good morning, and thank you for participating in Rigetti's earnings conference call covering the third quarter ended September 30, 2025. Joining me today is Jeff Bertelsen, our CFO, who will review our results in some detail following my overview. Our CTO, David Rivas, is also here to participate in the Q&A session. We will be pleased to answer your questions at the conclusion of our remarks. We would like to point out that this call and Rigetti's third quarter ended September 30, 2025 press release contains forward-looking statements regarding current expectations, objectives and underlying assumptions regarding our outlook and future operating results. These forward-looking statements are subject to a number of risks and uncertainties that could cause actual results to differ materially from those described and are discussed in more detail in our Form 10-K for the year ended December 31, 2024, our Form 10-Q for the 3 and 9 months ended September 30, 2025, and other documents filed by the company from time to time with the Securities and Exchange Commission.
These filings identify and address important risks and uncertainties that could cause actual events and results to differ materially from those contained in the forward-looking statements. We urge you to review these discussions of risk factors. During today's call, we will refer to certain non-GAAP financial measures. For details on these measures and reconciliations to comparable GAAP measures and for further information regarding the factors that may affect Rigetti's future operating results, please refer to yesterday's earnings release on Rigetti's website at investors.rigetti.com or to the 8-K furnished with the SEC yesterday after the close. Today, I'm pleased to report that during this past quarter, we saw strong momentum with both the demand for our on-premises quantum computers and the development of collaborations to advance our own R&D and the quantum ecosystem more broadly. On the technology front, we remain on track to deliver our 100-plus qubit chiplet-based quantum system with an anticipated 99.5% median 2-qubit gate fidelity by the end of 2025.
I'm also excited to share our 2026, 2027 road map updates. We expect to deploy a 150-plus qubit system by or around the end of 2026, with an anticipated 99.7% median 2-qubit gate fidelity. And by or around the end of 2027, we expect to deploy a 1,000-plus qubit system with an anticipated 99.8% median 2-qubit gate fidelity. In September 2025, we announced purchase orders totaling approximately $5.7 million for 2 9-qubit Novera quantum computing systems. Both systems are upgradable, allowing the customers to increase the system qubit count for more complex computations and research. One system is being purchased by an Asian technology manufacturing company. The system will serve as a testbed to develop internal quantum computing expertise. They also plan to benchmark and validate their own quantum computing technologies with the Novera system. The other system is being purchased by a California-based applied physics and artificial intelligence startup.
The system will be used for quantum hardware and error correction research. Our open and modular architecture continues to allow us to integrate innovative solutions with our technology stack, including our project with QphoX and the Air Force Research Laboratory or AFRL, to advance superconducting quantum computer networking. In September 2025, we announced a 3-year $5.8 million contract from AFRL to advance superconducting quantum networking. Rigetti will be collaborating with QphoX on the project, a Dutch quantum technology start-up developing leading frequency conversion systems for quantum applications. A key challenge to networking superconducting quantum computers is the need to convert the microwave signals, which are used to control superconducting qubits, to optical photons that can travel along those fibers. This project aims to deliver systems providing entanglement between superconducting qubits and optical photons, the essential building block of quantum networking.
Our new collaborations with the Center for Development of Advanced Computing, or C-DAC, and Montana State University showcase the increasing maturity of the quantum computing ecosystem. MSU is the first academic institution which has on-premises Rigetti quantum computer in 9-qubit Novera QPU, which will be used by researchers to advance quantum computing R&D. We intend to work with MSU on a variety of initiatives, including research projects related to quantum hardware and hybrid quantum systems, and co-development and testing of enabling quantum technologies and quantum system components. Collectively, these initiatives underscore the importance of public-private partnerships in advancing next-generation quantum technologies. We also signed a memorandum of understanding with C-DAC, India's premier R&D organization of the Ministry of Electronics and Information Technology. With this MOU, Rigetti and C-DAC intend to collaborate on the design and development of hybrid quantum computing systems and related technologies and bring them to market.
We are proud to be deepening our support for quantum computing capabilities in the academic and government sectors. We are equally excited to support NVIDIA NVQLink, NVIDIA's new open platform for AI supercomputer quantum integration. By providing low latency and high throughput integration between quantum hardware and AI supercomputing, NVQLink is a very promising resource to accelerate hybrid computation development on the path towards quantum advantage. We remain engaged with the Defense Advanced Research Projects Agency, or DARPA, on stage A of quantum benchmarking initiative or QBI project. On November 6, DARPA announced the companies initially selected to participate in phase B of the QBI project. Although we were not selected at this time for Phase B, we received constructive feedback regarding our proposal and we will continue to work with their team. We are optimistic that we will be chosen for Phase B in the coming months.
Lastly, I'm also pleased to share that Rigetti plans to open an Italian subsidiary in the coming months. We believe that this development will allow us to accelerate our pursuit of business opportunities and talent in Italy as the region dedicates more resources and funding to bolstering its quantum initiatives. Thank you. Jeff will now make a few remarks regarding our recent financial performance.
Thanks, Subodh. Revenues in the third quarter of 2025 were $1.9 million, compared to $2.4 million in the third quarter of 2024. On a year-over-year basis, our revenue for the quarter was impacted by the expiration of the National Quantum Initiative and its pending reauthorization in the U.S. Congress. Renewal of the U.S. National Quantum initiative sales to U.S. and foreign governments and Novera are all important to future sales. The recent sales Subodh noted in his remarks, the 2 9-qubit Novera system sales and the AFRL contract will benefit revenue in the fourth quarter and as we move into 2026. Gross margins in the third quarter of 2025 came in at 21%, compared to 51% in the third quarter of 2024. The lower gross margins on a year-over-year basis were due to the composition of our revenue and variability in the pricing in terms of our contracts. Our recent contracts with the U.K.'s National Quantum Computing Center for quantum systems have a lower gross margin profile than most of our other contracts.
On the expense side, total OpEx in the third quarter of 2025 was $21 million, compared to $18.6 million in the same period of the prior year. The increase in total OpEx was due to annual salary increases, new hires and higher stock-based compensation and consulting costs, primarily in research and development. Stock compensation expense for the third quarter of 2025 was $4.3 million, compared to $3.4 million for the third quarter of 2024. Our operating loss for the third quarter of 2025 came in at $20.5 million, compared to $17.3 million in the prior year period. Our GAAP net loss for the third quarter of 2025 was higher than our GAAP net loss for the third quarter of 2024 primarily due to the noncash change in the fair value of our derivative warrant and earn-out liabilities. We recorded a $10.7 million or $0.03 per share non-GAAP net loss for the third quarter of 2025, compared to a $13.4 million or $0.07 per share non-GAAP net loss for the third quarter of 2024.
As of September 30, 2025, we had approximately $558.9 million of cash, cash equivalents and available-for-sale investments and no debt. Subsequent to September 30, 2025 and through November 6, 2025, proceeds of $46.5 million were received from the exercise of slightly more than $4 million of our public warrants. As of November 6, 2025, cash, cash equivalents and available-for-sale investments totaled approximately $600 million. Thank you. We would now be happy to answer your questions.
分析師問答
Our first question comes from David Williams of The Benchmark Company.
Maybe first, Subodh, just kind of thinking about the DARPA Phase B, and just can you talk maybe a little bit about that? You said that you've received some nice or constructive feedback. But can you maybe talk around what is maybe holding that up and when you think we might have an answer or you might see that advancement happen?
Sure, David. So as we mentioned in the press release, DARPA did the initial selection of companies that they have got into Phase B. Unfortunately, we were not one of them, but they gave us good constructive input on what we need to do and what to improve on to get into Phase B. So we are working on that. And it primarily goes into the area of error corrections and some areas of long-range coupling, things that are important in the long term to get to a DARPA fault-tolerant quantum computing milestone in 2033. Not as important in the short term to get to quantum advantage. So a lot of our focus has been and continues to be on getting to quantum advantage in the next 3 to 5 years with 1,000 qubits and 99.9% 2-qubit gate fidelity and with some error correction. DARPA's input was more on the FTQC milestone and where we need to increase effort further, specifically in the area of error correction and long-range coupling.
So we are incorporating that input. We will continue to talk to DARPA. We are still a very much part of Phase A and we'll continue to work with DARPA closely. So we're optimistic we'll get into Phase B soon. Exactly when, that's hard to know, but we'll continue to work on it. But I mean, DARPA's project, as you know, is a 7-year project. So just because we didn't make the initial cut, it's not a big deal. We feel pretty good that we'll make the cut in the next few months here.
Okay. Great. It seems like this is more of a conceptual discussion rather than about actual performance or what you are currently achieving, but in the long term, how would you describe some of the performance metrics? Is that accurate?
That's fair to say. Fundamentally, the data is really good, and we are very proud of what we have demonstrated with our Ankaa's system, as well as the 36-qubit chiplet-based system that has 99.5% 2-qubit gate fidelity and about 69 to 70 gate speed. That data is impressive and positive. The constructive criticism, however, revolves around how we will handle error correction and long-range coupling to achieve the FTQC milestone 7 to 8 years from now. It’s really about the future work that our plan needs to improve upon. So, it's a reasonable observation, as you pointed out.
Great. And then maybe just secondly, I think in the past you mentioned having over 1,000 qubits, 99% fidelity, and around 50-nanosecond gate speeds to achieve quantum advantage. Looking at your roadmap toward 2027, you're quite close to that, though possibly a bit short on fidelity. So my question is, how confident are you that you can achieve 99% fidelity by 2027? Is that a reasonable target for when you anticipate reaching quantum advantage, or do you believe that timeframe might be pushed further out?
No, it's a good question. And really, we are excited to disclose that the 2 big milestones, one for '26, we believe we will hit 99.7% fidelity at the 150-plus qubit level. But more importantly, the 2027 milestone when we believe we will get over 1,000 qubits at 99.8% 2-qubit gate fidelity. You're right, I mean, it's a significant jump-up from where we are, and frankly, the whole quantum computing industry is, including peers in superconducting quantum computing, but certainly when you look at other modalities, those numbers are impressive with 1,000 qubit, 99.8% at 16-nanosecond gate speed. It gets us awfully close to quantum advantage, but not quite there. For quantum advantage, we still think we need a 99.9% 2-qubit gate fidelity, as well as some form of error correction. So between '27 and '29, which is when we still believe we accomplish quantum advantage, is getting the fidelity to that 99.9% and also error correction up. Hopefully, that answers your question.
It does.
Our next question comes from the line of Quinn Bolton of Needham & Company.
Subodh, Jeff, I wanted to follow up on David's question about the roadmap to 150 qubits next year and over 1,000 by 2027. Subodh, could you explain whether this will still be a chiplet-based approach? Will it utilize 9 qubit tiles, or as we aim for the 1,000 qubit system, do you anticipate increasing the number of qubits per tile? Additionally, considering DARPA's interest in quantum error correction and long-range coupling, can long-range coupling be achieved with the tile-based system? What are your thoughts on this?
Sure, great questions, Quinn. We're planning to use 9-qubit chiplets for the 150 qubit system with 99.7% two-qubit gate fidelity. For the 1,000 qubit goal, we’re currently considering scaling up to 36-qubit chiplets to reach that milestone at 99.8% two-qubit gate fidelity by the end of 2027. We’re confident in this plan due to the chiplet architecture and the data we're collecting from our current 36-qubit system, along with experiments on 100-qubit systems that we aim to launch soon. In terms of DARPA's focus on error correction and long-range coupling, we haven't encountered any particular challenges with chiplets or long-range coupling. The issues are essentially the same whether using a single monolithic chip or a chiplet-based system. Long-range coupling is a challenge for the entire industry, not just us, and we've found that chiplets don’t complicate that issue. The main concern remains the physical distance between qubits and how to effectively couple them. We do need to address long-range coupling as part of our development, but the presence of chiplets doesn’t exacerbate the challenge. I hope that clarifies your question.
That's great. And then one for Jeff. Jeff, I think you mentioned in your script, the AFRL contract as well as the 2 9-qubit Novera sales would start to generate revenue in the fourth quarter and into 2026. I guess maybe on the 2 Novera sales, I think you, in the press release, talked about completion or delivery of those systems in the first half of 2026. Is this sort of a revenue recognition that you'd be able to recognize those sales upon delivery because their systems, maybe not just QPUs, is there a percentage completion accounting that is used for those systems? Maybe just walk us through how you recognize revenue on the Novera sales if they're systems rather than just QPUs?
Sure. On the 2 Novera system sales, I mean, we anticipate recognizing the revenue for those upon shipment. Right now, it looks like one of them will go in the first quarter, one in the second quarter. But upon shipment would be the manner of revenue recognition.
Our next question comes from the line of Krish Shankar of TD Cowen.
This is Steven speaking on behalf of Krish. My first question is for either Subodh or Jeff regarding the two Novera system sales you mentioned. I'm curious if both orders are complete systems that include dilution fridges and full control systems, or if one of them might just be a QPU chip. Additionally, regarding the upgrade option, is it included in the price you announced, or is it an additional revenue step that would come later?
Sure. I'll take that. So the 2 systems, they include everything from dilution refrigerator to control systems. So they're complete systems. Regarding upgrade, when the customers upgrade them from 9-qubit to, let's say, 36-qubit or something bigger, it will be an additional revenue opportunity because we have to go and add some cables and those kinds of things inside the dilution refrigerator to account for the additional qubits. Certainly, obviously, the chip has to change too. So there will be additional revenue that comes with the upgrade from 9-qubit to a higher qubit count sometime in the future.
I wanted to ask about the future support for NVIDIA's NVQLink interface. Can you discuss the software or hardware changes that will be needed for your QPUs or control systems to support it? Additionally, do you have any thoughts on hybrid quantum computer support? Is this primarily for the supercomputing space, or could NVQLink also facilitate the integration of quantum systems alongside AI data centers for GenAI applications?
Great question, Steven. So if you look at NVIDIA's NVQLink announcement, it's an open format for quantum computers to basically interface directly with AI supercomputers. So the idea is indeed to have quantum computing start being used with GenAI and potentially for AGI type applications. Now from our viewpoint, this was a natural step. We have always said that we believe in the hybrid systems, we have always supported hybrid standards. And that's partly because of the strength of superconducting quantum computing that we have. Speed is around commensurate with CPU and GPU speed. So it's logical for us to try to interface with HPCs. And that's why we believe superconducting quantum computing is most amenable for hybrid computing compared to other modalities which are 1,000 times slower, like trapped ion or pure atom modalities. So for us, it was a logical step when NVIDIA started discussing an open platform like NVQLink.
We obviously signed up with it. It fits in with our vision and strategy of having a quantum computer as part of a hybrid ecosystem. We certainly expect products like that to start coming into data centers once we get closer to quantum advantage, although interfaces will be worked out between now and then. So the timeline for having quantum computers in data centers doing practical applications doesn't change because of the NVIDIA announcement. What it does change is the whole notion of how a hybrid system will work and open standards that support the hybrid systems. Hopefully, that answers your questions.
Our next question comes from the line of Craig Ellis of B. Riley Securities.
I wanted to follow up on a couple of prior questions to start. So Subodh, with regard to NVQLink, NVIDIA is very, very strong in the National Labs. Rigetti has a very strong position in National Labs. So can you talk about what Rigetti's historic strengths with National Labs mean for engaging with ecosystem partners that can help accelerate Rigetti's integration with hybrid compute and getting pulled into various workloads, including AI-related workloads with NVQLink?
Sure, Craig. So you are indeed right, I mean NVIDIA has a very strong presence in national labs, and so do we with quantum computing. So it's logical for the interfaces to be worked out at National Labs level, whether it's for the National Lab or the Oak Ridge National Lab or other national labs. Also the NQI initiative, although not funded at the higher level, the funding has restarted last week, as you probably saw. So it's exciting to have National Labs get their funding back again to some reasonable level and this NVQLink platform being launched at about the same time period. So certainly, we believe, as we have discussed in the past that in future, CPUs will continue to be used for sequential computing and GPUs will be used for parallel computing as they are being used today. And QPUs, quantum processing units, will be used for simultaneous computing. So everything we have discussed in the past, now we have a chance to start demonstrating it in real life in partnership with NVIDIA, with their NVQLink platform as well as the quantum platform.
So definitely expect more work in this direction where we will be able to generate data, where we will be able to take generic applications and split them into sequential parallel and simultaneous and show how the three respective technologies are suitable and the benefit of having the three technologies work together in a complementary way. That we believe is the best way to address future computation needs.
That's really helpful, Subodh. And Jeff, I wanted to ask a follow-up clarification to you. Regarding the AFRL deal at $5.8 million, I think that was a 3-year deal for $5.8 million. Does that revenue recognition occur fairly ratably across 12 quarters? Or how do we think about revenue recognition? And is that kicking up in the fourth quarter or early next year?
No. It will be fairly distributed over the three years, Craig. We actually recognized a portion of it in the third quarter, so it will be consistent going forward.
Our next question comes from the line of Brian Kinstlinger of Alliance Global Partners.
A follow-up on the road map. I'm curious what progress you are making currently on fidelity and when you expect to achieve 99.7% medium 2-qubit gate fidelity for a 9-qubit chip and when that has to happen in order to start the tiling process to get to 100 qubits by the end of 2026?
We are currently working on 9-qubit chiplets and aiming to achieve our milestones for this year, targeting over 100 qubits with a fidelity of 99.5% before the year ends. We are observing strong 2-qubit fidelity with the 9-qubit chiplets, giving us confidence that we can reach 99.7% by the end of next year with more than 150 qubits. The 1,000 qubit goal presents more challenges. We plan to increase the chiplet size to around 36 qubits, and we need to validate that we can connect multiple chiplets while maintaining high fidelity. This will be our focus next year as we aim to demonstrate over 1,000 qubits at 99.8% by the end of 2027. The data we are collecting from the 9-qubit chiplet gives us strong confidence in achieving this year’s goal of over 100 qubits at 99.5%, as well as next year's target of over 150 qubits at 99.7%. The chiplet data supports our confidence in meeting both of these milestones, which is crucial for reaching 1,000 qubits at 99.8% by the end of 2027.
Our next question comes from the line of Richard Shannon of Craig-Hallum Capital Group.
Subodh and Jeff, let me ask a couple of questions here. Looking at your 10-Q, and you have a passage in here about you may significantly increase your CapEx, including upgrading our chip fab facility or an entirely new one here. Maybe you can tell us a little bit about what this potential might be? When you might decide this? And what's the kind of scale of investment we're talking about here?
Right now, we have a 150-millimeter chip fabrication facility in Fremont, California, which operates in a mostly manual manner. It is currently providing us with good data, and we are confident it will continue to meet our needs for the next two to three years, including achieving the 1,000 qubit at 99.8% milestone by the end of 2027. However, the challenge lies in reaching over 99.9% 2-qubit gate fidelity with several thousand qubits. We believe the current facility will face capability limitations, not due to capacity but because the tools designed for 150-millimeter chips do not perform as well as those for 200 or 300 millimeters, particularly since the semiconductor industry has largely standardized around those sizes. Therefore, we anticipate needing 200 and 300-millimeter tools and more automation for increased capability beyond the next three years. Typically, constructing a new fab requires a couple of years, so if we need a new setup in three years, we will likely need to consider capital expenditure about a year from now.
There are discussions about various alternatives involving commerce and national labs within the U.S., and if those initiatives move forward, we will participate rather than shouldering the full investment for an 8-inch or 12-inch fab alone. Quantum fabrication is significantly simpler than a cutting-edge CMOS fab because our lateral dimensions are more forgiving, and our primary challenges are vertical dimensions relating to oxidation. Additionally, we have far fewer lithographic steps compared to a CMOS fab. This means establishing a new quantum fab, whether 8-inch or 12-inch, would cost hundreds of millions of dollars, whereas a new CMOS fab can cost between $20 billion to $25 billion due to the complexity and lateral dimensions involved. Therefore, a quantum fab is fundamentally much cheaper in comparison to a new CMOS fab, though still involves sizable investments in the hundreds of millions. This reflects what was mentioned in the 10-Q, indicating we may begin exploring this option if there is no national initiative led by commerce or others that allows us to join in.
It does, Subodh. Let me follow up on that topic here, specifically regarding the balance between having something stand-alone versus shared, as well as the concerns around intellectual property and potential leakage. What are the considerations involved in making that decision?
Puts and takes are similar to what we see in the regular semiconductor industry. Currently, the most advanced fabs are operated by foundry models like TSMC, which manage IP security very effectively. Companies like NVIDIA and AMD are producing their advanced chips at TSMC without any IP issues. Foundries have successfully learned how to meet different customer design requirements while maintaining IP protection. If a foundry model becomes established in the U.S., where a state-of-the-art fab does not currently exist, it would be beneficial for us to explore this option since it has proven successful. However, even if this doesn’t materialize, the financial implications are manageable, with potential costs in the hundreds of millions rather than tens of billions, as seen in CMOS. Therefore, it is possible for us, either independently or in partnership with other companies, to pursue these developments without completely adopting a foundry model.
Okay. That's helpful detail here. Maybe a follow-on question here for Jeff. A number of questions here earlier in the call here about the future revenue recognition for both the AFRL contract as well as the system sales here. How do we think of kind of a general profile of gross margins of these additional revenues are? Or Kind of general thought process here, especially since gross margins here in the third quarter are lower than what you've seen in the past.
Sure. So gross margins were lower, as you pointed out in the third quarter. Really is due to the variability in our contracts, and sometimes we do these contracts for strategic reasons or because they're going to advance our R&D necessarily more than the margin profile. I do think with some of these other sales, particularly some of the Novera sales, margins will be a bit better than, certainly, than what we've seen here in the third quarter and even earlier in the year to a certain extent.
Our next question comes from the line of Troy Jensen of Cantor Fitzgerald.
Congrats on all the great progress here. Maybe a couple of quick questions for Subodh. Just curious on the 2027 target of 1,000 qubits, what types of applications would your system be able to run at that status?
Great question, Troy. This is where I think the exciting part comes in. The announcements we've made with NVIDIA, NVQLink, and hybrid systems are all coming together around the same timeframe. Imagine a world where there is a hybrid system offering involving us, NVIDIA, and a few other companies, featuring 1,000 physical qubits with 99.8% 2-qubit gate fidelity, smoothly interfacing with state-of-the-art CPUs and GPUs. We believe the types of applications you will be able to tackle will be the complex ones that struggle with CPUs and GPUs today. We are talking about things like drug discovery, financial forecasting, and material synthesis. We don’t think we will be focusing on encryption or decryption at that point with those metrics. However, applications with thousands of variables that interact simultaneously will begin to come our way as current CPU and GPU architectures struggle to keep up. As I mentioned, many forecasting-type applications, whether financial forecasting, drug discovery, or weather forecasting, are the ones I expect will start utilizing quantum computing in a hybrid setup in a couple of years.
Perfect. All right. And then just a question, I'm just curious here. The customers that are buying these 9-qubit systems, why would they not buy the 36-qubit system now?
Good question. We asked the question to them, too. And they are buying physical on-premise quantum computers right now because they want to fundamentally understand how quantum computers work, because they are doing some research on some aspect of quantum computing themselves. But they need to understand how fundamentally the hardware works, what kind of pulses do how do we send, how do we tune, recalibrate and retune and all those kinds of things. So 9-qubit is a good starting point for those kinds of things to understand how a quantum computer works. But as we discussed, they clearly are interested in upgrading it at the right time. Once they are confident they understand how 9 qubits work, I'm pretty sure they will want to upgrade it to tens of qubits, whether it's 20-odd qubits or 30-odd qubits, we'll see what their interest level is. But they will certainly be interested in upgrading. And that's why the systems are designed so that they are upgradable. There will be an additional revenue recognition at that point because we've to change the chip, we've to change the wiring and a few other things. Fundamentally, the systems are designed so they can handle up to 50-or-so qubits.
Our next question comes from the line of David Williams of the Benchmark Company.
Just wanted to ask Subodh, if you kind of think about your foundry as you spoke about earlier, is there a possibility that you could transfer your technology today to an outside fab that's slightly more advanced that you could get better fidelity? And just kind of thinking about the 1,000 qubit, is there a potential to maybe get to that 99.9% alternatively using another fab source?
Good question, David. And we are talking to existing foundries that are doing some quantum computing-related work for some of our peer companies. So I mean, we are exploring those options. And certainly, if it gives us some bump up in performance, we would love to have it. As of today, we haven't found anyone who's quite that capable of running the types of materials and processes that we run for superconducting gate-based quantum computing. We know in the other forms like superconducting annealing and some other modalities like photonics, there are some foundry companies that are doing some work for some of our peer companies. And we are talking to them to see if we can use that model. But as of right now, all the leaders in superconducting gate camp, including us and other companies like IBM and Google, as far as I know, none of us are using a foundry model at this time. But we will continue to explore those options. If an existing foundry meets our requirements, we would love to have it obviously. It saves us a lot of CapEx, if that is the case. But as of today, we are not confident that the existing foundries can meet our requirements.
Okay. And then maybe just one last one here on Craig's question about M&A earlier. Do you think you have the right kind of path forward on the control side? And you kind of talked about needing to transition to a flexible cabling platform. Is that an area that you could potentially be looked to outside sources for acquiring? Or do you think you have that under control today and have the path forward there?
Well, I mean, on the control system itself, as you know, we are partnered with Quanta Computer who is a leader in CPU/GPU servers. So we feel very good about our strategic partnership with Quanta Computer for the control system itself. Regarding the cables that go inside the dilution refrigerator, you're right, we need to move to flex cables in the next 2 to 3 years. We have good technology ourselves along with some subcontractors that we use right now. We have a lot of IP in that area as well. So we feel generally good about our path forward. But if, again, as I mentioned, if we find someone who can help us accelerate our road map, we will absolutely be willing to take a look at that kind of a company.
Our next question comes from the line of Quinn Bolton of Needham & Company.
Subodh, you mentioned the Energy Department announcing, I think it was $625 million to invest in the National Quantum Research Center. Just wondering how does that affect the business? And do you have any updates what you're hearing in Washington on just the reauthorization of the NQI Act?
So yes, the original NQI was signed in 2018 and ran out of funding towards the end of 2023. The reauthorization was expected to be signed at that time, but it hasn’t happened yet. There have been many discussions with various versions of the bills moving through the House and Senate, with proposals as high as $2.5 billion over 5 years, which is about 4 to 5 times more than the original NQI funding of $625 million. The latest version that passed last week has reinstated the original funding amount, bringing it back to $625 million over 5 years, or $125 million per year. While it's better than having no funding, it falls short of the $2.5 billion that has been discussed. This is just the first step, and ongoing discussions suggest there could be a much better funding situation for DOE in the next few months. However, given the current government circumstances and the timeline for signing and appropriating bills, we cannot specify the exact date or amount. We do expect a significantly larger funding discussion for DOE in the coming months, but it's positive to see us back to the original amount of $625 million over 5 years.
Our next question comes from the line of Tyler Anderson of Craig-Hallum Capital Group.
This is Tyler Anderson on for Richard Shannon. So I have a housekeeping and a technical question. For Q4, what do you guys expect the share count to be? I noticed there was a few warrants that were exercised subsequent to the end of the quarter.
Sure, Tyler. So I would say it's going to depend on how many warrants get exercised, of course, between November 6 and the end of December. As of November 6, we had 330 million shares outstanding. So I would probably plan on maybe 335 million or something like that. Again, it could flex a little bit depending on how many warrants get exercised.
Okay, great. Regarding the tools you plan to acquire for the new foundry, are you waiting for new tooling to be developed, or will the tools you expect to buy be upgradable as new capabilities become available? If they are upgradable and you plan to enhance them, how does that affect your roadmap in terms of qubit density and fidelity?
The current road map we disclosed for 2026 and 2027 still relies on our Fremont facility to provide us with the necessary chips. We do not plan on needing a new fabrication facility or foundry to meet our goals. We believe the milestones for 2026 and 2027 are achievable with our existing fab. We are considering options that include current foundries, and if a foundry model is viable, it will be the simplest solution to implement. If not, we will explore the possibility of investing in a new fab, either independently or through initiatives from the U.S. government. We are open to various approaches. Any new fab would likely be either 8-inch or 12-inch, and it's improbable that most of our current tools could be reused, except for minor items. Therefore, most equipment would need to be newly acquired at the 8-inch or 12-inch scale, which represents a significant capital expenditure. However, relative to CMOS costs, we're still talking about smaller amounts, in the hundreds of millions of dollars, assuming the existing foundry model fails and we have to shoulder the full responsibility ourselves. We hope to discover more straightforward and cost-effective solutions.
Partially. So with the tools that you purchased, would those potentially be upgradable for just quantum add-ons that tooling companies are thinking about?
Yes, they should be. I mean, fundamentally, they are very similar kind of tools, so they should be upgradable in the future.
And is there any capability that you would look for in the foundry, if you were to purchase one?
I mean right now, as you know, in the superconducting gate camp, we use what is called as Josephson junctions, and then we create gates between the qubits. So materials themselves are superconducting materials like aluminum, tantalum and those kinds of materials, which are not normally available in the CMOS world. The processes, there are some unique processes that we do to enable our superconducting gate chips. Again, so slightly different materials and slightly different processes, that's what we need. Some foundries that are doing like superconducting annealing type approaches, they have some of the tools that we need, but not all of them. So those are the options we are looking at right now, to see whether we can use some of those models after our Fremont fab before we have to commit to a brand-new fab ourselves.
Congrats on the roadmap.
Thank you. I'm showing no further questions at this time. I'd now like to turn it back to Dr. Subodh Kulkarni for closing remarks.
Thank you for your interest and questions. We look forward to updating you after the end of next quarter. Thanks again.
Thank you for your participation in today's conference. This does conclude the program. You may now disconnect.