管理層發言
Good day, ladies and gentlemen, and thank you for standing by. Welcome to the Rigetti Computing First Quarter 2026 Financial Results Conference Call. As a reminder, this conference call is being recorded. At this time, I would like to turn the conference over to Mr. Subodh Kulkarni, CEO of Rigetti. Sir, please begin.
Good afternoon, and thank you for joining us for Rigetti's First Quarter 2026 Earnings Conference Call. I'm pleased to be joined today by our Chief Financial Officer, Jeff Bertelsen, who will walk you through our financial results in more detail following my overview. Also with us is our Chief Technology Officer, David Rivas, who will be available to participate in the Q&A session following our prepared remarks. We appreciate your continued interest in Rigetti, and we look forward to answering your questions at the conclusion of our remarks. Before we begin, I would like to remind everyone that today's call, along with our first quarter 2026 press release, contains forward-looking statements. These statements reflect our 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 anticipated. Such risks and uncertainties are described and discussed in greater detail in our filings with the Securities and Exchange Commission, including our Form 10-K for the year ended December 31, 2025, our Form 10-Q for the three months ended March 31, 2026, and other periodic reports filed by the company from time to time with the SEC. We encourage you to review these filings for a comprehensive discussion of these risks and uncertainties that could cause actual events and results to differ materially from those contained in the forward-looking statements. Rigetti undertakes no obligation to update any forward-looking statements made during this call, except as required by law. 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 today's earnings release on Rigetti's website at investors.rigetti.com or to the 8-K furnished with the SEC today after the close. Before I begin, I want to frame today's discussion around three key takeaways. First, with the general availability of our 108-qubit Cepheus-1-108Q system on Rigetti Quantum Cloud Services, Amazon Braket, Microsoft Azure Quantum and qBraid, we believe we have delivered one of the most powerful generally available quantum computers in the world and the largest modular quantum computing system on the market today. Second, we are seeing growing adoption of Rigetti systems across government, academic and commercial customers, including new on-premises Novera QPU sales that support meaningful year-over-year revenue growth. Third, we remain focused on disciplined execution against our roadmap to quantum advantage: continuing to improve Rigetti on Cepheus-1-108Q and advancing toward higher-qubit, higher-fidelity chiplet-based systems underpinned by a strong balance sheet and prudent capital deployment. Now I'll step back and put the quarter in context. Q1 was an important proof point in our strategy to combine technical progress with real-world access and usage. Quantum computing remains a long-cycle opportunity, but we are increasingly seeing the ecosystem coalesce around platforms that can scale in a practical way and that are available to users where they already run their workloads. Our progress this quarter reflects that reality. Let me start with our technology and product milestones. Last month, we announced the general availability of our 108-qubit Cepheus-1-108Q quantum computing system accessible to customers via Rigetti Quantum Cloud Services and through Amazon Braket as well as Microsoft's Azure Quantum service and qBraid. Cepheus-1-108Q is our highest qubit count system to date and the industry's largest modular quantum computing system built from 12 interconnected 9-qubit chiplets. This system triples the number of qubits and chiplets from our previous 36-qubit Cepheus-1-36Q system, and more importantly, validates some proprietary chiplet-based scaling architecture in a production setting. Today, Cepheus-1-108Q has achieved a median two-qubit gate fidelity of approximately 99.1% with gate speeds of roughly 60 nanoseconds and a median single-qubit gate fidelity of 99.9%. These are meaningful performance levels at this scale, and we expect to continue improving fidelity throughout 2026 as we refine the performance of our individual chiplets, innovate across materials and fabrication and incorporate learnings from our prototype and R&D platforms. We achieved a median 99.8% two-qubit gate fidelity with 40-nanosecond gate speeds on our 9-qubit system by using a proprietary adiabatic CZ gate scheme. Leveraging the same gate scheme, we also demonstrated two-qubit gate fidelity as high as 99.9% at 28-nanosecond gate speeds on a prototype system, and those advancements are informing how we operate Cepheus-1-108Q and design future systems. From a systems engineering perspective, this launch is about more than just adding qubits. During development, we identified and mitigated coupling interactions between tunable couplers that become more pronounced beyond the 100-qubit scale. By defining our chip architecture to address those interactions, we effectively shifted the primary performance limitation from coupler behavior to coherence time, which we are confident we can address as we continue to optimize our entire stack. Also I want to highlight what this means for users. With Cepheus-1-108Q now available on Rigetti QCS, Amazon Braket, Microsoft Azure Quantum and qBraid, researchers and enterprises can access our highest qubit count system on platforms they already use for classical and quantum R&D. Cepheus-1-108Q is the first gate-based device on Amazon Braket with more than 100 qubits, offering improved fidelities that enable wider and deeper circuits for applications such as material science, optimization and quantum simulation. AWS is the leader in cloud infrastructure, so extending our relationship with Amazon Braket and now Azure Quantum and qBraid is an important validation of our technology and our go-to-market strategy. Stepping back, we continue to believe that superconducting gate-based quantum computing with chiplet-based scaling offers a compelling combination of speed and scalability. Our current systems achieved gate speeds on the order of 50 to 70 nanoseconds, which is roughly 1,000x faster than some alternative modalities such as trapped ion or neutral atom systems. As we scale, we intend to maintain those speed advantages while driving fidelity higher and integrating error-correction-ready operations into the stack. Let me now turn to customer momentum and market traction. Our strategy is to meet with customers where they are, whether that is on the public cloud, on hybrid infrastructure or in dedicated quantum centers. On the cloud side, the combination of Rigetti QCS, Amazon Braket, Microsoft Azure Quantum and qBraid provides global access to our systems, including Cepheus-1-108Q, and we are seeing strong interest from researchers who want to experiment on one of the most capable generally available gate-based platforms in the market today. In parallel, we continue to expand our base of on-premises Novera QPU. The Novera QPU is designed to integrate into a customer's existing cryogenic and control systems, providing a high-performance, on-premises platform for quantum R&D. Recent Novera events include an order from the University of Saskatchewan, where our QPU will support quantum research and education. And we have also announced Novera QPU and Novera system sales to additional research organizations globally. These systems deepen technical engagement, create multiyear usage pathways and showcase the flexibility of our product portfolio from 9 to more than 100 qubits. As discussed in our prior call, Novera and other system deliveries contributed to significant year-over-year growth, albeit with some variability quarter-to-quarter based on shipment timing and contract mix. For example, we expect a meaningful portion of previously announced Novera purchase orders to be recognized in the first half of 2026, and we are executing on additional system-level contracts such as the C-DAC order we announced earlier this year. While the timing of revenue recognition can move between quarters, these contracts underscore growing demand for Rigetti QPUs and systems among national labs, universities and quantum computing centers. We are also encouraged by continued engagement from commercial customers who are exploring quantum-inspired and hybrid use cases. While commercial revenue remains early, we are seeing increased interest from industries such as materials, logistics and financial services as they look to understand where quantum computing can augment classical high-performance computing over time. More broadly, we are starting to see tangible examples of how even relatively small-scale quantum systems can impact real-world workloads. For example, a team in China recently demonstrated that a 90-bit quantum system could outperform classical reservoir networks with tens of nodes on a realistic weather forecasting task, highlighting how modest-sized quantum devices can begin to disrupt AI and modeling applications. We view results like this as early validation of the commercial opportunities systems like our Novera QPUs and Cepheus-class devices are positioned to address as they mature. Let me briefly connect this back to our long-term roadmap. We remain focused on a clear sequence of milestones that we believe positions Rigetti to reach quantum advantage in roughly three years. Near term, that means driving Cepheus-1-108Q to a median two-qubit gate fidelity of approximately 99.5% later this year while maintaining our gate speed advantages. Beyond that, we are working towards deploying systems that leverage our chiplet-based architecture as the foundation for eventually scaling to more than 1,000 qubits with fidelities and gate speeds that support error-mitigated and ultimately fault-tolerant computation. In support of this roadmap, we recently announced our intention to invest up to $100 million in the United Kingdom over the next several years to accelerate quantum computing development. This will be our first major investment outside the United States and builds on our existing 36-qubit system deployment at the U.K.'s National Quantum Computing Centre as well as the U.K. government's multibillion-dollar commitment to quantum technologies. In parallel, we continue to collaborate with partners such as River Lane and others to integrate error-correction-ready capabilities into the stack. This includes support for high-fidelity native gates, improved noise-aware compilation and control electronics enhancements that are designed to be compatible with future error-corrected architectures. Our intention is to update our published technology roadmap later this year once we have incorporated operational data from Cepheus-1-108Q and can provide more detail on the specific steps we expect to take towards quantum advantage. Turning to the financial framework. Our approach remains straightforward and disciplined. We exited last year with a strong cash position and no debt, giving us the flexibility to continue investing behind our technology roadmap and customer opportunities. Our spending remains concentrated in core R&D, including fabrication, chip designs and control electronics development, along with the capital expenditures required to support higher qubit count systems and associated cryogenics infrastructure. While this results in elevated CapEx in the near term, we believe these investments are directly tied to the capabilities that will differentiate Rigetti in the market. We are not managing the business around short-term revenue optimization. We are managing it around credible progress towards large-scale, high-fidelity quantum systems that can deliver commercially meaningful value. To that end, our capital allocation remains focused on organic execution and we would consider M&A only where we can clearly accelerate our roadmap without compromising our financial discipline. To close my remarks before turning it over to Jeff, I want to reiterate the three key messages we hope you take away from today's call. First, Cepheus-1-108Q is now generally available through Rigetti QCS, Amazon Braket, Microsoft Azure Quantum and qBraid, and we believe it represents one of the most powerful generally available gate-based quantum computers in the world and the largest modular system on the market today. Second, customer adoption continues to build across cloud and on-premises channels with Novera sales and other contracts, supporting strong year-over-year revenue growth and deepening our engagement with leading research institutions and emerging commercial users. Third, we remain committed to disciplined execution on the roadmap that targets quantum advantage in about three years, anchored in our chiplet-based architecture, high-speed superconducting qubits, improving fidelity, a strong balance sheet and strategic initiatives such as our planned $100 million U.K. investment that enables us to invest with patience and control. Thank you for your continued support and interest in Rigetti. I'll now turn the call over to our CFO, Jeff Bertelsen, who will walk you through our financial results in more detail.
Thank you, Subodh, and good afternoon, everyone. I will spend a few minutes walking through our first quarter 2026 financial results, our balance sheet and how we're thinking about capital deployment as we continue to execute on the roadmap Subodh described. For the first quarter of 2026, revenue was $4.4 million compared to $1.5 million in the first quarter of 2025. The year-over-year increase was driven primarily by on-premises Novera QPU deliveries and related contracts as well as certain government and research projects. Gross margin for the first quarter was 31% compared to approximately 30% in the first quarter of 2025. Our first quarter 2026 gross margin was impacted by contract mix, including a higher contribution from QPU and system deliveries that include lower-margin third-party refrigeration. Total operating expenses for the first quarter were $27.3 million compared to $22.1 million in the same period last year. Spending remains concentrated in research and development, including engineering headcount, fabrication and system integration, consistent with the priorities we outlined on our fourth quarter call. Stock-based compensation for the quarter was $5.9 million compared to $4.2 million in the first quarter of 2025. Operating loss for the first quarter was $26.0 million compared to an operating loss of $21.6 million in Q1 2025. On a GAAP basis, net income for the first quarter of 2026 was $33.1 million compared to net income of $42.6 million in the prior year period. The first quarter of 2026 included $53.7 million of noncash gains from the change in fair value of derivative warrant and earn-out liabilities compared to $62.1 million in the prior year period. As a reminder, these noncash fair value adjustments can introduce significant volatility into our GAAP results quarter-to-quarter and do not affect how we operate the business or allocate capital. On a non-GAAP basis, which excludes stock-based compensation and fair value adjustments to warrant and earn-out liabilities, net loss for the quarter was $14.7 million or $0.04 per diluted share compared to a non-GAAP net loss of approximately $15.3 million or $0.05 per diluted share in the first quarter of 2025. Let me provide a bit more color on revenue drivers and how we are thinking about the remainder of the year. As we outlined in our fourth quarter call, we expected strong year-over-year revenue growth in the first quarter of 2026, driven by shipment of a portion of the $5.7 million of on-premises Novera quantum computing system purchase orders announced late last year. The first quarter results are consistent with that view, and we continue to expect the remaining Novera revenue to be recognized primarily in the second quarter of 2026. We also continued to execute on the $8.4 million C-DAC order for an on-premises 108-qubit system in India, which we expect to recognize in the fourth quarter of 2026 following installation and successful completion of acceptance testing. As we said last quarter, the initial C-DAC order did not include ongoing maintenance and support. We still expect to receive a separate purchase order for those services. More broadly, our revenue profile continues to be influenced by the timing of system deliveries and government-funded projects. We continue to view this variability as inherent to the current stage of the market and not as a driver of our long-term capital allocation or technology strategy. Turning to the balance sheet. We ended the first quarter of 2026 with approximately $569 million in cash, cash equivalents and available-for-sale investments compared with $209.1 million as of March 31, 2025, and approximately $589.8 million at the end of 2025. The year-over-year increase relative to Q1 2025 reflects the capital raise and strategic investment activity we have previously discussed, while the sequential decline from year-end reflects ongoing operating spend and capital expenditures. We continue to operate with no debt. At our current operating profile, we believe our capital position provides sufficient runway to execute against the technology and system deployment milestones we have laid out, including continued progress on scale, fidelity and system integration as well as our planned investment in the United Kingdom. Capital expenditures in the quarter were primarily driven by investments in Fab-1 and additional dilution refrigeration capacity to support higher qubit count systems over the next several years, consistent with the framework we outlined in the fourth quarter. We continue to expect 2026 CapEx to be elevated relative to prior years, largely due to refrigeration and infrastructure needs rather than major changes to our fab footprint. Our approach to capital deployment remains disciplined and consistent with what we have discussed on the Q4 call. The majority of our spending is directed toward core R&D activities that directly advance our technology platform, including our chiplet-based architecture, control systems and cloud integration. We are not managing the business around short-term revenue optimization. We are managing it around long-term progress toward quantum advantage and commercially relevant systems. To close, our financial strategy is unchanged from what we outlined last quarter. We are focused on maintaining flexibility, funding innovation responsibly and aligning capital deployment with the long-term value-creation potential of our technology roadmap. While quarterly results will continue to reflect the early-stage nature of the quantum computing market and the timing of large system contracts, we believe our balance sheet and capital discipline position us to execute with patience and control. With that, I will turn it back to the operator, who will open the call for your questions.
分析師問答
Our first question or comment comes from the line of Brian Kinstlinger from Alliance Global Partners.
I'll ask two. The first is: can you talk about the announced NVIDIA quantum models, when you expect they might be available and when you might begin to test them to see the impact they have on reducing your error rates?
Thanks, Brian. NVIDIA did announce an open-source model called NVIDIA Async to help with calibration and bring-up of quantum computers as well as error-correction tooling. We continue to look at that as a possible means of accelerating our roadmap. We continue to talk to NVIDIA, and we also continue to talk to other partners in the industry, such as River Lane in the U.K., where we are partnering to do error correction. These are not replacements for each other; they can work in a complementary fashion. So certainly, the announcement made by NVIDIA to help accelerate quantum computing in terms of calibration, bring-up and also error correction is something we are taking a close look at, and we'll definitely take advantage of those tools that are available now. Hopefully that answers your question.
Yes. Great. My follow-up: the $100 million investment in the U.K. — is that primarily for people, infrastructure, offices? And then will that be expensed or capitalized? If it's expensed, when will we start to see that begin to increase OpEx?
So let me put that U.K. investment in context. The U.K. has announced a fairly ambitious program that they call ProQure, which is a multistage program. Right now, the first phase will kick off this July or August for a couple of years; the next phase that the government will kick off at that time runs for another year or two, and then GigaQuOp and so on. MegaQuOp means one million error-free quantum operations per second; GigaQuOp means one billion error-free quantum operations per second. It's a very well-structured program right now; applications are being requested. We will be one of the applicants. Assuming we are chosen for the preliminary phase, we definitely plan to increase our headcount, so there will be additional personnel costs. We definitely plan to increase the number of quantum computers we have in the U.K. Right now, if you visit the National Quantum Computing Centre outside Oxford in the U.K., you will find our quantum computer in that center. For the next phase, we definitely plan to include Cepheus-1-108Q or higher-qubit-count quantum computers there over the next couple of years as we make them available. So there will be some capital costs involved but also facilities. Right now, our quantum computer sits in the NQCC facility; we currently have a relatively small office, and we definitely plan to have a larger facility in the U.K. as we go forward. So the $100 million is over the next few years and captures the rough order of magnitude of those costs. Hopefully that answered your question.
Our next question or comment comes from the line of Krish Sankar from TD Cowen.
I just wanted to ask you on the integrated error mitigation: is this an on-chip qubit gate, or is it a separate control chip? Is it ASIC or FPGA? Can you give us some color on that? Got it. And just a question on the QPU pipeline: how is that looking, how has it evolved? Is the funnel expanding — besides the U.S. government — is it stable or how should we think about it?
Krish, are you talking about the NVIDIA-specific announcement? In general, right now, we don't do error correction on the quantum chip itself. There are approaches being explored to do that, but our quantum chip is not performing error correction at the chip level today. Most of the error correction and the experiments we do are outside in the control systems area: we send signals to the quantum computer from the control systems and get the responses back from the quantum computer. Regarding the QPU pipeline and demand: interest in quantum computing continues to increase rapidly. As we start getting closer to what we define as quantum advantage — roughly a 1,000-qubit system with 99.9% two-qubit gate fidelity and gate speeds under about 50 nanoseconds, and some form of error mitigation or correction — we roughly think that's about three years from now. We are already starting to see increased interest from not only academic and national-lab customers but also commercial customers who want to do quantum computing-related R&D activities. We expect that interest to continue to increase rapidly as we approach those milestones, and we're already starting to see that reflected in our disclosures and sales numbers. Overall, we expect quantum computing interest to increase even though we remain in R&D stages right now.
Our next question or comment comes from the line of Quinn Bolton from Needham and Company.
A follow-up on Krish's question: with the adiabatic CZ process that you're already showing on prototypes getting to 99.9% two-qubit gate fidelity, how long does that take to get into production processes? It sounds like you're targeting 99.9% as part of the system that gets to quantum advantage. Why does it take so long to get there, and what are the steps needed to bring that process from prototype level into higher-volume production?
It's a good question, Quinn. We will obviously push as fast as possible to get adiabatic CZ and fast gates into higher-scale systems. We are already using adiabatic gates in our Cepheus-1-108Q as well, but it's not the very fast adiabatic CZ gate that we have been able to achieve at the prototype stage. At the prototype stage, we had 99.9% with gate speeds of 28 nanoseconds. So you can see that our 108Q is still slower than our prototype system and fidelity is not as high as the prototype system. We take the learnings from the prototype system and try to include those improvements in our larger-scale systems as soon as possible. It just takes time — these are extremely complex problems to solve at scale. It's relatively easier to demonstrate performance on prototypes, which is why many announcements from various organizations are at the sub-10-qubit level. When you start getting to 100 qubits or above, the engineering challenges increase significantly. There are only a few companies that have enabled quantum computers at that scale, and we are proud to be one of them with the 108Q system that is available for anyone to use. The problems become more significant to tackle and solve as you scale. We have demonstrated very strong performance with adiabatic CZ and fast gates at small qubit counts; we need to take those learnings and push them as fast as we can. They will be part of the systems that approach quantum advantage in three years, and I expect to see adiabatic CZ and fast gates included in systems we deploy next year and the year after.
Got it. And then, as you look to that system that gives you quantum advantage, do you expect that system to run quantum error correction? Or would you still be thinking about implementation of on-chip quantum error correction being beyond the quantum advantage chip in roughly three years?
It's a good question. There's still a lot to be determined regarding quantum error correction. We discussed earlier how tools like NVIDIA Async could influence and potentially accelerate our roadmap. Our view right now is that the quantum advantage system — roughly a 1,000-qubit system at 99.9% two-qubit gate fidelity — will use some form of error mitigation and possibly limited error-correction techniques, but not full, large-scale quantum error correction. We envision full, large-scale quantum error correction, including schemes like quantum LDPC (low-density parity-check) codes, to be part of a later phase of fault-tolerant quantum computing, which would require hundreds of thousands of qubits. We think full fault-tolerant quantum computing is in the roughly five- to seven-year timeframe. Quantum advantage will involve substantial error mitigation and partial error correction techniques, but full implementation of fault-tolerant quantum error correction is further out.
That does. I appreciate it. And then maybe just one quick one for Jeff. You said on the $5.7 million of Novera QPU sales that you had announced last year, you expected to capture most of the remainder in the second quarter. Looking at the 10-Q you filed today, it looks like you had about $3 million of hardware-based sales. So is the remainder of that $5.7 million, roughly $2.7 million, to be recognized in Q2? Is that the right ballpark?
Yes. Of that $5.7 million, we recognized a little bit less than half of that in Q1. We expect the remainder to be recognized in Q2.
Our next question or comment comes from the line of Craig Ellis from B. Riley Securities.
I want to start following up on some of the comments you made about the 108-qubit Cepheus availability on Rigetti QCS and then on Amazon Braket and Microsoft Azure Quantum and qBraid. As it's generally available, what are you seeing in terms of engagement across the various platforms? And are you getting any feedback in terms of what the workload tests are shaping up to be?
Good question, Craig. It's still relatively early to talk about usage because we only deployed the system about a month ago. Interest is high, and we are seeing significant usage, but it's too early to draw firm conclusions based on the early data. We expect usage to increase over the next few months as more researchers and enterprises become aware of the system and begin to explore and test workloads. We will continue to improve the system's fidelity, and we expect to deploy a higher-fidelity 108Q configuration later this year. As performance improves, we expect usage to increase further.
That's helpful. As a follow-up, you established the Quanta partnership in Q1 of 2025 and the investment in Rigetti was formalized in early Q2 2025. As you look back at the first year of that deal, what would you identify as the top two or three things that are really going well and helping you scale up qubit count and system capabilities? And what would be one or two things you would hope the partnership could do this year?
Thanks, Craig. We entered into a strategic partnership with Quanta. They invested about $40 million in Rigetti at that time, but more importantly there was a commitment on both sides to collaborate and invest across the hardware stack. One of the key accomplishments so far is how well they have designed and delivered a new control system that we have started including in our more recent offerings. Our latest deployments to customers are exploring the use of Quanta-made control systems instead of entirely home-built systems. Quanta is a large, experienced company with deep capabilities in servers and hardware, so they bring professionalism to building control-system boxes that meet our requirements. They have a dedicated team working on systems that integrate with our platform. Going forward, we expect to continue using Quanta's control systems as part of our stack. It's not exclusive — we will continue to maintain our capabilities in that area and Quanta can work with others as well — but we are clearly benefiting from their expertise in control systems. Over the next year or two, we expect them to continue improving control systems to meet our system requirements and to contribute to other parts of the hardware stack.
Our next question or comment comes from the line of Vijay Rakesh from Mizuho.
On Cepheus-1-108Q, any thoughts on what you're getting for price uplift versus the 36-qubit system? And how has customer response been since making Cepheus-108Q available on Azure and Braket? Second, on the C-DAC $8.4 million win: when do you expect that to layer into revenue and when do shipments occur?
Vijay, it's still early; we deployed the system just over a month ago, so it's early to quantify usage uplift versus 36Q. Interest is high and many customers are using the system now, but most usage at this stage is research-focused with short-duration jobs — on the order of seconds to a few minutes — as people experiment and develop algorithms. We're not yet seeing large-scale commercial, data-center-style jobs; we expect that to ramp as we approach quantum advantage over the next couple of years. Cepheus-1-108Q is one of the most powerful cloud-accessible quantum computers available today, so we expect interest to continue to grow as performance improves. Regarding the C-DAC order, as we stated when we disclosed the order, our plan is to fulfill that order in the second half of this year, most likely in Q4. Installation, acceptance testing and commissioning are part of that timeline, and we expect most of the associated revenue to be recognized when the system is installed and accepted before year-end.
Our next question or comment comes from the line of Antoine Legault from Wedbush Securities.
Could you remind us how you remain confident that the architectural fix you recently achieved with couplers is durable as you scale beyond 108 qubits to a few hundreds and eventually over 1,000 qubits?
That's a good question, Antoine. We look at this carefully as we update our roadmap. Our fundamental architecture continues to be a square grid with tunable couplers — an approach similar to other major superconducting efforts. Tunable couplers give an extra degree of freedom to adjust coupling between qubits, which is important for managing frequency placement and interactions. Where we diverge is our focus on chiplet-based scaling: it's fundamentally easier to build smaller chiplets than a large monolithic chip, and that makes scaling more practical from a fabrication and yield perspective. We have designed our chiplet architecture to address the coupling interactions that emerged at larger scale, and so far we are not seeing architectural constraints that would prevent scaling with tunable couplers and chiplets. We feel confident that our roadmap from the current 108 qubits — improving fidelity and increasing qubit count using chiplets — is solid and executable toward our quantum advantage target.
Our next question or comment comes from the line of Troy Jensen from Cantor Fitzgerald.
You said quantum advantage is about three years away for you. Is it fair to say that's roughly two chip cycles away? And can you talk about chip cycles: previously you were spinning chips roughly every nine months; will that slow going forward to 12 to 18 months? Any color would be helpful.
Troy, as you know, we operate our own fab. Chip-cycle timing depends on the scope of changes we make. We typically launch major revisions about once a year, but because we control the fab we can iterate faster when focusing on specific improvements. We can turn around revisions faster than once a year if required — potentially twice a year for certain changes. Three years gives us plenty of major iterations toward quantum advantage. A big part of the path to quantum advantage is chip improvements, but remember there are many components across the stack — dilution refrigeration, cabling, control systems, error mitigation, software — that all contribute to final system performance. So while chip cycles matter, it's a broader system integration effort. We think three years is a realistic timeline to get the metrics in place for quantum advantage.
You mentioned dilution refrigeration. That's an area where competitors without dilution refrigeration could press you on cost. Can you talk about what needs to happen there and, once you're at quantum advantage, what the dilution refrigeration costs look like compared to alternative modalities?
In superconducting quantum computing, we do need dilution refrigeration to cool chips down to millikelvin temperatures for superconductivity. Dilution refrigeration technology has existed for decades in specialized applications and is now being industrialized for quantum. There are several commercial suppliers of dilution refrigerators — for example, Bluefors and Oxford Instruments — and we have relationships with multiple suppliers and track their roadmaps closely. For scale to thousands of qubits, the refrigerator form factor and integration roadmap we see from suppliers supports the qubit counts we plan to reach; the qubit chip area even at 108 qubits is small, and at several hundred to a thousand qubits the dimension is still modest. Dilution refrigeration is complex, but it's not a fundamental bottleneck for superconducting platforms. Other modalities don't require such cryogenics, but superconducting qubits bring large advantages in speed and circuit performance — gate speeds on the order of tens of nanoseconds, which can be 1,000x to 10,000x faster than some room-temperature modalities. Those speed and scalability benefits make the cryogenics tradeoff acceptable. We will continue to monitor refrigeration suppliers and integrate improvements as they become available.
Our next question or comment comes from the line of John McPeake from Rosenblatt Securities.
Subodh and Jeff, congrats on getting Cepheus out on the cloud. I have a couple of questions. You're saying later this year you expect to get to 99.5% from 99.1% — could you dig in on what needs to happen? I have a quick follow-up after that.
Sure, John. As we disclosed, the limiting factor for two-qubit fidelity right now is coherence time — the time we can maintain quantum states. Currently our coherence times are in the 25 to 30 microsecond range. We aim to roughly double or ideally triple those coherence times to move from around 99.1% toward 99.5% median two-qubit gate fidelity. We know the mechanisms affecting coherence and have specific experiments and process improvements underway in fabrication, materials and packaging to improve coherence. So we feel confident we can improve coherence and therefore fidelity over the course of the year.
The lab machine at 99.9% — was that a 9-qubit system or what physical qubit count was used for that result?
The highest-fidelity prototype data were achieved on a very small prototype; our recent demonstration at scale referenced a 9-qubit system that reached 99.8% with 40-nanosecond gates using the adiabatic CZ scheme. The 99.9% data were observed at the prototype level with faster gates. We need to carry those techniques to 36 qubits and then 108 qubits; that's the engineering challenge we're actively addressing.
Any update on DARPA? I thought you were going to reengage with them.
We continue to be engaged with DARPA as part of their program. They provided feedback late last year and we are working on those items: error correction, scaling challenges and related milestones. It's an open-ended program with milestone-based advancement into subsequent phases. As we hit milestones, we expect to progress into Phase B and beyond. We remain engaged and are making progress on the items they identified.
Finally, at the end of the year I think there was discussion of a 150-qubit machine. Should we still think about that?
Yes. Our primary long-term milestone is quantum advantage in about three years. Right now we're at 108 qubits and low- to mid-99% fidelities. As we scale, there will be times when we increase qubit count without increasing fidelity and times when we improve fidelity without increasing qubit count. Ideally we do both, but it's often staged. This year we expect to introduce a higher-fidelity 108-qubit system and we are evaluating 150 qubits or higher as we iterate. We're focused on trying to increase both qubit count and fidelity as rapidly as practical.
Our next question or comment comes from the line of Richard Shannon from Craig-Hallum Capital Group. We will not be able to pull up Mr. Shannon's line at this time. I would like to turn the conference over to Mr. Subodh for any closing remarks.
Thank you for your interest in Rigetti's earnings call and for the thoughtful questions and discussions today. We are encouraged by the progress we are making on our technology roadmap, the growing engagement we are seeing from customers across cloud and on-premises channels, and the strength of our balance sheet to support disciplined execution. We remain focused on delivering against the milestones we have laid out and on building a business that can create durable long-term value as quantum computing matures. On behalf of the entire Rigetti team, thank you for your continued interest and support, and we look forward to updating you on our progress next quarter.
Ladies and gentlemen, thank you for participating in today's conference. This concludes the program. You may now disconnect. Everyone, have a wonderful day. Speakers, stand by.