Good afternoon, and welcome to Inflection's third quarter 2026 earnings conference call. Thank you for joining us today. My name is Marcus Kuperschmidt, Head of Investor Relations and Strategic Finance. Joining me today are Matt Kinsella, our Chief Executive Officer, Panav Gokhale, our Chief Technology Officer and General Manager of Quantum Computing, and Alon Hart, our Chief Financial Officer. Following our prepared remarks, we will open the call for questions. A replay of this call will be posted on our Investor Relations website, along with our earnings press release and the presentation materials. Before we begin, I would like to remind you that this conference call will include forward-looking statements, including statements regarding our future financial condition, business strategy, and plans and objectives of management for future operations, and all other statements that are not statements of historical fact. These statements are subject to risks, uncertainties, and assumptions that could cause actual results to defer materially. These factors are detailed in the earnings press release we issued today and other filings with the SEC, including our annual report on Form 10-K for the year ended December 31, 2025, which are available on our website at ir.inflection.com. We undertake no obligation to revise or update any forward-looking statements, except as required by law. During today's call, we will also reference certain non-GAAP financial measures. We use these measures because we believe they provide additional insight into the underlying operating performance of the business. This non-GAAP financial information should not be considered in isolation or as a substitute for GAAP results. Reconciliations between GAAP and non-GAAP measures can be found in today's press release and in our SEC's filings.
With that, I'll turn the call over to Matt. Thank you, Marcus, and good afternoon, everyone. Whether you've been with us since we became public or you're dialing in for the first time, welcome, and thanks for your interest in inflection. In May, I described Q1 in one word, momentum. In August, the word for Q2 was acceleration. This quarter, the word for Q3 is execution, specifically technical execution. We committed to demonstrating 30 logical qubits by the end of 2026, and we delivered ahead of schedule. And as many of you have heard me say before, logical qubits are the keys to the kingdom for useful quantum computing. Achieving 30 logical qubits on only 80 physical qubits is a major milestone to getting us to utility-scale quantum computing, and Pranav will go into far more detail on those results. In addition to our technical execution in compute, we are seeing strength in the demand environment in our computing business. Customer deployments and other leading indicators, such as growth in backlog and revenue performance obligations, RPOs, give us increasing confidence that compute revenue will grow strongly in 2027 and will represent a meaningfully larger share of our total revenue. Revenue in the quarter was $9.1 million, up 25% year-on-year, and down quarter-on-quarter, and Alon will walk through this in more detail, but the sequential decline and associated gross margin impact reflects two specific government programs and the revenue recognition timing associated with them. We expect this to reverse in Q4, and we are maintaining our full year revenue guidance of approximately $45.1 million. Now, since we aren't strangers anymore, I'll keep the company overview brief this time. Then I'll talk a little bit about why I'm so excited about what quantum can make possible, and then touch on some of the progress we made this quarter. And as I mentioned earlier, Pranav will then take you through our progress in computing in much more detail. I'll jump back in and cover sensing. Alon will review the financial results in more detail, and then I will close with our priorities before we move on to Q&A. So just a quick refresh on inflection. We build quantum computers and quantum sensing systems on a shared neutral atom hardware and software platform. Nature gives us our qubits in the form of atoms, and our job is to turn those atoms into working systems by controlling them with light and integrating the hardware and software around them. And because our products share the same core capabilities, progress in one area strengthens the others. And we're the only company in the quantum industry combining this breadth of products on a single platform, giving us multiple paths to commercial value. All neutral items, all the time. Let me just give you some examples of the problems our customers are trying to solve and what excites me most about quantum. And it's really, it's potential to move entire industries forward. And only a handful of technological breakthroughs in history have fundamentally changed what's possible. And when they do, that impact is extraordinary. If we just think about clocks, for instance. For centuries, sailors struggled to know exactly where they were at sea. Accurate marine clocks helped solve that problem that completely transformed navigation and opened up a new era of global trade. Or take the transistor. That one invention laid the foundation for modern computing and eventually put a powerful computer in every one of our pockets. And finally, radio. For the first time, we could communicate across vast distances without wires, and it changed how the world fundamentally communicated. I believe quantum can have as big or bigger of an impact than those prior technical revolutions, and at inflection, we have opportunities to make that kind of impact across our entire portfolio. Quantum computing and quantum sensing give us new ways to model and measure the physical world from the materials that we build with and the energy that we produce to the capabilities that our national security depends on. So let me start with computing, and let's take materials as an example. All materials are made of atoms, and the interaction between atoms-electrons determine many things that range from how a battery stores energy to how a semiconductor conducts electricity to how a display produces light. But simulating those interactions between those electrons exactly becomes exponentially more demanding each time an additional electron is added. So even a molecule with a relatively small number of atoms can make simulating the interaction of those electrons far beyond the reach of today's most powerful classical computers. The problem is truly intractable for the digital compute architecture. Quantum computers operate using the same laws of physics that atoms do, which opens up the possibility of modeling these interactions in ways classical digital computers cannot. And advances in hardware, error correction, logical qubits, and algorithms are bringing us closer and closer to using that capability to designing better materials, developing new medicines, and improving how we produce and store energy. And making these fundamental breakthroughs in our ability to produce new materials will be absolutely game-changing for humanity. And the opportunity extends beyond materials to areas like energy and finance. In energy, we're working with customers to improve grid reliability. And in finance, our customer is building new approaches to modeling asset pricing. For applications, this scale, even small improvements can create significant economic value. And Pranav, again, we'll walk through these applications and our progress in computing in a little bit. And now if we consider sensing, in some ways, we've come full circle to that marine chronometer. Today, GPS relies on satellites equipped with atomic clocks. Your phone knows where it is by precisely timing the signals from those satellites. When GPS is jammed or unavailable, navigation and precise timing can be compromised, and our atomic clocks can provide an independent source of precise time when GPS is disrupted, and we're developing inertial sensors that can help ships and aircraft navigate by measuring their own movement without relying on satellites. And if we think about RF, radio revolutionized our ability to send information through the air. Quantum spectrum has the potential to revolutionize our ability to listen to and understand it. By using atoms themselves as sensors, we have the opportunity to change how we detect and locate radio transmissions, particularly in crowded or contested environments where conventional technologies face limitations. We're also advancing quantum gravity sensors for space. While exploring how that same technology could help us see beneath the Earth's surface, identifying underground structures associated with critical minerals, and giving customers better information before they drill. The stakes are enormous. Global economic competitiveness and our national security. And we're building inflection to lead in both computing and sensing. So let me turn to our progress this quarter. This quarter, we announced collaborations with NVIDIA, Cisco, and Riverlane, and in Japan, our quantum processing unit now powers the country's first operational full-stack neutral atom quantum computer, and we were the only foreign quantum partner selected for Japan's quantum moonshot program, putting our technology at the heart of this national initiative. We launched our Quantum Application Forum series in London with a highly successful first event, drawing interest from enterprises across financial services, energy, pharmaceuticals, consumer goods, aerospace, and defense. And what impressed me was the range of companies exploring quantum computing and the business problems they want to solve. That breadth of interest reinforces our confidence in the commercial opportunity ahead. NASA expanded our contract by $20 million to support a mission designed to fly the world's first space-based quantum gravity sensor, bringing the total contract value to $40 million. And in Quantum Spectrum, we announced an approximately $12 million contract from the UK government in October to advance our direction-finding technology toward a deployable product. We saw accelerated growth in our backlog and remaining performance obligations, fueling continued momentum for 2027 revenue growth across both computing and sensing, and we expect ongoing growth in those metrics during Q4. And given the progress we are making in computing, we are increasingly confident that computing will represent a meaningfully larger share of our revenue next year. We increased investment in our teams and capacities to support growth, particularly in computing. We laid out those plans on our last earnings call, and we're following through with them. These investments will help us move faster and deliver the platforms and technologies that we need to win. All right, Pranav, we'll now walk you through our 30-logical qubit result and why it matters, what it tells us about our approach, and where we're going from here. Pranav, over to you.
Thanks, Matt. We're delighted to share more about the completion of our flagship RedMap target for 2026, achieving 30 entangled logical qubits with our scale-neutral atom quantum computer. Let me first distill why this milestone is a significant step forward. In our demo, all 30 logical qubits worked together in a computation validating the coordinated performance of our hardware and software, and we achieved this with just 80 physical qubits. Fewer than three physical qubits per logical qubit, reinforcing our confidence that we can achieve our future targets with less hardware than originally anticipated. We also demonstrated error correction techniques that recover results on a 30-logical qubit state that would otherwise be discarded, reducing the time needed to obtain usable results. Together, these advances support our next milestone, 100 logical qubits, capable of 1 million reliable operations, or 1 mega-quap, in 2028. As we progress, we will increasingly emphasize both the number of logical qubits and the number of quantum operations they can reliably perform, which are two complementary measures of our progress toward useful quantum computing. I'm now going to get more technical. Logical qubits are the singular pathway to useful quantum computation. Just as your Wi-Fi router or your 5G phone translates noise-prone packets of radio signals into pristine streams of useful data, useful quantum computers will translate noise-prone physical qubits into pristine logical qubits. Our demonstration sets the commercial neutral atom record for the largest ever entangled logical state. I'll dig in on the next slide with the key takeaways. We exceeded both technical targets for our 30 logical qubit circuit that I outlined in our Q2 earnings call. First, we had targeted advanced circuits using fewer than 10 physical qubits per logical qubit, and ultimately, we achieved fewer than three physical qubits per logical qubit. That efficiency reflects an AI-driven advance that reduced the number of physical gates. Needed for logical entanglement, as well as our ability to move atoms to connect distant pairs of qubits as needed, referred to as all-to-all connectivity. Second, we targeted an atom loss rate below 2% per entangling gate, and we ended up outperforming that target by a factor of 2. This is especially important for inflection's approach with neutral atoms, in which we intentionally convert hard-to-identify errors into atom loss. A missing atom tells us exactly where a problem occurred, making it easier for our software to recover the error-corrected information. Our signature demonstration ran an algorithm on 30 logical qubits using approximately 1,000 physical operations, or 1 kiloquat. Running circuits of this size, we evaluate the fidelity of the entire application or computation, and even after accumulating errors across all 1,000 physical operations, our error detection experiment achieved a circuit-level fidelity of 21% across all of the operations, whereas random guessing would give a success rate of just 0.02%, giving us a 1,000x signal-to-noise ratio. In addition, when we error correct for atom loss events, we boost the system yield by a factor of 36, while only reducing application-level fidelity by a factor of 3. We will release a paper with our technical details within the next few weeks. In addition to this signature demonstration, we executed another key building block for reliable computation by preparing 30 logical qubits in a tesseract quantum error correction code. Captain America fans may remember the tesseract, the powerful glowing cube, and ours won't open a portal to another universe, but it does help protect quantum information, which for our roadmap is considerably more useful. This code has an error protection metric, known as distance, of 4, meaning that it can be used to correct any single physical error. In our state preparation experiment, we tracked 15 X stabilizers and found a per logical qubit success rate of about 99.9% with error correction for atom loss and 99.2% per logical qubit with error correction for bit flips. Most importantly, these error correction procedures boost the usable data by 3x and 11x, respectively, over error detection. These results demonstrate our ability to correct errors while retaining more usable data, a capability we will build on as we extend the duration and complexity of our computations. Our next step is to increase both the number of logical qubits and the number of quantum operations they can reliably perform. For investors, these are complementary measures of progress. How much quantum data the machine can store and how much work it can execute. Reaching our 2028 target of 100 logical qubits and 1 mega-quop will bring together advances This is across our hardware and software. The number of quops captures how underlying hardware capabilities, including physical gate fidelity and physical qubit count, combine with compiler and error correction software to determine how much computation the system can reliably perform. Through this hardware-software co-design, we now expect to achieve mega-quop capability with as few as 2,000 physical qubits, a significant improvement in efficiency relative to prior estimates. This efficiency informs both our R&D and our design for manufacturability, supporting our capacity to build QPUs in volume. Our megaclop system will integrate internal advances inflection has made in continuous atom reloading with constant magnetic fields and repeated syndrome extraction for long-duration error correction. These capabilities will allow us to replenish atoms. and repeatedly identify and correct errors to sustain large computations. These technologies also support our roadmap toward 1,000 logical qubits in 2030. That brings me to what customers are doing with our technology. In Japan, our QPU customer, the Institute for Molecular Sciences, Announced that the Shunkai machine is now operational as the nation's first full-stack neutral atom quantum computer. At the Illinois Quantum and Microelectronics Park, we are finalizing site readiness as we prepare to deliver the QPU in 2027. And the customer pipeline for new hardware engagements continues to build. Engagement with customers on applications is increasingly shifting to co-development. Starting that work now allows customers to help shape the applications and the systems they will ultimately run on. In the energy sector, our customer Eaton is developing solutions for how quantum computing can assist with grid contingency analysis and outage planning, an application with both economic and national security impact. Imagine a utility preparing for a summer heat wave. Air conditioning demand is surging, and a major transmission line is already offline for maintenance. The utility needs to know if another line or generator fails, can the remaining grid carry the load, or could that trigger a cascading blackout? The challenge is evaluating the many combinations of equipment failures, demand levels, and available generation. A grid might withstand each individual failure, but become vulnerable when several occur together. Our work with Eaton is exploring how quantum computing can help utilities evaluate these complex scenarios more effectively, the potential values identifying vulnerabilities earlier, giving utilities better information to take preventative measures before an outage occurs, leading to fewer disruptions and a more resilient grid. Now, moving to the finance sector, a leading global financial institution and inflection customer ran tens of thousands of shots of a 30-logical qubit circuit on our scale QPU in Q3. The exciting results validate ongoing investment to build more powerful quantum-based models for financial asset optimization by evaluating probability distributions of complex financial scenarios that will become intractable for classical computers to evaluate. Our customer has a very large asset base, so providing even small improvements relative to current classical modeling techniques equates to significant sums of money. We continue to see growing customer interest in our computing capabilities, and we continue to see onboarding of additional customers who are eager to build with us now. Back to Matt's throughline of execution, in the past quarter, Inflection executed on our computing roadmap to deliver more logical qubits and more capable qubits, advancing us, ultimately, toward more useful applications. And we are continuing to execute with intense focus in the coming quarters as we prepare for 100 logical qubit systems with 1 million quantum operations approaching in 2028. And with that, let me turn the call back to Matt.
Thank you, Pranav. And let me turn over to the sensing side of our business. Sensing opens large markets for quantum technology, from national security and critical infrastructure to space. And we have built a broad portfolio to serve those markets, spanning precision timing, inertial sensing, radio frequency sensing, and gravity sensing. Our focus is on growing commercial adoption and moving more and more of that portfolio into deployable products. In precision timing, we recently demonstrated Ticker aboard a U.S. Navy support vessel, where it operated continuously under demanding conditions at sea. And we're also adding new customers for national security use cases. Our software helps customers identify what matters in large, complex sensor data sets, and our quantum-inspired AI runs on today's GPUs, supporting work with the U.S. Navy, U.S. Army, and the European Space Agency. And let's walk through two recent customer contracts, and we'll start with Quantum Spectrum. The UK government has awarded us a contract worth approximately $12 million to accelerate the productization of quantum spectrum. Quantum spectrum represents the first fundamental shift in RF sensing architecture in decades. Conventional systems often need separate antennas and receivers for different frequency bands with some antenna arrays that are as big as a football field, and that can put parts of the spectrum out of reach for smaller or mobile platforms that need those signals for communications, navigation, and threat detection. With quantum spectrum, our atom-based antenna can be as small as a sugar cube, and the system can cover a wide range of frequencies with less receiving hardware. We're seeing growing demand from defense customers and are investing in the performance, reliability, and testing needed to bring deployable products to market. And beyond defense, we're seeing potential in compact receivers for long-range communications. Direction finding could also help locate RF interference at airports or other critical infrastructure and help with search and rescue teams to find distress signals. In quantum gravity, NASA has expanded our contract by $20 million, bringing the total contract value to $40 million. The mission, led by NASA's Jet Propulsion Laboratory, or JPL, aims to fly the world's first quantum gravity sensor in space. The expansion of our next phase of hardware development and testing, with work expected to continue through 2027, and we will put our quantum core through microgravity testing at the Einstein Elevator in Germany, which is a critical step preparing the sensor to operate in orbit. We are developing the technology for future quantum sensing systems to track changes in water and ice from space, so changes in aquifers and the polar ice caps. Here on Earth, quantum gravity sensing can help map geological structures associated with critical mineral deposits, groundwater, and identifying underground hazards. Representative Jeff Hurd's bipartisan bill would direct the U.S. Geological Survey to use quantum gravity sensing to improve mineral mapping, giving companies better information before they drill. All right, Alon will now take you through the quarter in more detail. Alon, over to you.
Thanks, Matt, and good afternoon, everyone. I will walk you through the highlights of our gap and non-gap result for Q3 2026. Revenue was $9.1 million, up 25% year-over-year, with all revenue organic and entirely from quantum. For the first nine months of 2026, revenue totaled $32.6 million, up 57% year-over-year. Approximately 85% of Q3 revenue came from the United States. As we have noted, geographic mix will vary with the timing and location of program awards. Our Q3 2026 gap operating loss was $34.5 million, compared with $6.7 million a year ago. Q3 2026 non-GAAP operating loss was $22.7 million compared with $5.9 million a year ago. The year-over-year increases primarily reflects investment in headcount and technology development, as well as the incremental cost of operating as a public company. The increase in GAAP operating loss also reflects higher stock-based compensation expenses. Q3 2026 revenue and gross margin declined sequentially, reflecting a timing of revenue recognition on the government program following an expansion award during the quarter, as well as lower margins on the final stages of another contract as it concluded. The magnitude of this impact was specific to Q3, and we expect revenue and gross margin to improve in Q4. Cash used in operations during Q3 2026 was $16.3 million, and capital expenditures were $5.2 million, compared to Q3 2025, level of $6 million and $0.1 million, respectively. Increasing investment primarily supported computing headcount to advance our roadmap, along with expanded facilities and labs to support expected growth. We continued to expense all R&D costs as incurred. We ended the quarter with $560 million in cash, cash equivalents, and available for sales security, down $22 million quarter over quarter. Our quarter-end cash balance included $28 million temporary working capital benefit from payroll taxes collected on stock options exercise. We are maintaining our 2026 guidance for a modest uptick in cash burn, which remains low relative to peers. We are deploying capital strategically to support our long-term objectives. I want to provide a brief update on the Department of Commerce transaction announced during Q2. We expect to complete a definitive agreement process later this year. Under the current structure, we expect that up to $100 billion will be recorded as a financing transaction. Final treatment will depend on the definitive agreement and the applicable accounting review. Back to you, Matt.
Thank you, Alon. And let me close where I began. The word for Q3 was execution. We committed to 30 logical qubits in 2026, and we delivered ahead of schedule. We are building for the long term, and each milestone gives us a stronger foundation for the next milestone. As our computing systems begin solving commercially relevant problems for customers and our sensing business grows, we expect revenue to grow faster than our operating costs, and that's how we'll build operating leverage over time. And the strength we're seeing in computing is particularly encouraging. Growth in our backlog and remaining performance obligations accelerated in Q3, and we expect both to continue growing in Q4. And that gives us greater confidence in our growth next year, with computing representing a meaningfully larger share of revenue. And I'll just say 2027 is shaping up to be what I call a target-rich environment. Marcus, back to you. Thanks, Matt.
Before we open up the call for questions, a quick note on where you can find us over the next several months. For those of you who want more specifics, please see our Investor Relations website. So let's go to Q&A, some of the ones we've aggregated before we get to the Southside Q&A. So first question, Matt, the opportunity for quantum spectrum. Is that primarily about improving existing customer capabilities, enabling new applications? What do you think matters most in terms of moving to the next phase of technological maturation?
Well, the magic of quantum spectrum is that it fundamentally rethinks how we receive radio frequencies, and in particular, which I've said many times, it allows you to receive very long wavelength, low frequency signals at very small form factors where normally you need a very big antenna. So that's kind of the lowest hanging fruit new application. And the range of possibilities and what that can enable are very wide. But the one that this new contract in the UK is specifically doing is replacing huge, huge, huge antenna arrays that can be the size of a football field with something that is very small that you can hold in your hand to do direction finding. And that is helping you identify things of interest that might be coming at you that you want to know about. So that is the application that we're focused on now. But the range of new applications is very wide, and it's really all brought about by the fundamental rethinking of how we receive radio frequencies. And the last thing I'll say is normally when you're receiving. You're emitting, almost always with classical RF technology. In this case, that is not the case. We are receiving entirely passively, and that is a huge implication in the battlefield.
That's great. Thank you, Matt. Pranav, second question that came in. The computing roadmap was always couched in logical qubits, and now we're hearing about QOPS. Can you help us understand the difference and what really matters?
Yeah, sure. So both are important. And in fact, there's a very natural analogy to classical computing, where in a classical computer, your laptop, you care both about the memory capacity and about the processing power. So for instance, when you go into Best Buy, and you buy a 32 gigabyte RAM computer, you sure don't want a Pentium II processor from decades ago on that machine, you wanted that memory to be And so it's the same for quantum computing. And specifically for us, for instance, as we get towards 100 logical qubits of capacity of quantum storage space, it's important that we pair that with one mega quap or one million quantum operations so that we can actually put those 100 logical qubits to good practical use for end customers. So that's the motivation for expressing both, and they are complementary metrics that matter in tandem. Thank you, Pranav. All right. So one more before we go to live to the Southside analysts. One more question in terms of the quantum software work and the designs that we're doing. What are the customer's main objectives and milestones? How do customers assess if this software is helpful if the hardware is not yet available to run on it? Sure. So I would say at the highest level, the way we think about this is under the core principle of ensuring that our customers can hit the ground running on our quantum hardware. And that's why the software-hardware co-design is so important, because the software work can begin months, if not years, before the hardware is ready. And maybe just to give you a few examples, I think the first, which is one of the shining selling points of the Neutral Atom platform, is that we have reconfigurable connectivity or qubits. And that means that we effectively have the FPGA-like technology of the quantum computing world where we can highly customize applications to the underlying hardware the same way that AI inference, cryptocurrency mining, network routing have been highly customized to FPGA flexible hardware. So that is one of the dimensions in which quantum software co-design can help bring out 10x, 100x, 1000x advantages over what you do with just default operations and how we help our customers around running. And then the other two tidbits I'll throw out there is resource estimation. This is where we've open sourced tooling that helps our customers understand how big a machine, how many shots of execution they need, how long the runtime is going to be. That's another way that our customers can figure out which applications are going to be best suited for which quantum computer and at which time frame. And the final piece is that this is. We're intricately linked to simulation, which is another area that we've pushed the frontier on how much we can effectively have a digital twin of our hardware before that hardware arrives. And this is all, again, under this core organizing principle of making sure that our customers hit the ground running. That's the implication of hardware software co-design. Thanks.
Operator, can you please open the line for questions?
Operator
Thank you. And just a note to the audience, if you'd like to ask a question, press star 1 on your phone. A confirmation tone will indicate your line is in the question queue. You may press star 2 if you would like to remove your question from the queue. For participants using speaker equipment, it may be necessary to pick up your handset before pressing the star keys. And your first question comes from Jesse Bolson with U.S. Bancorp BTIG. Please state your question.
Hey, everyone. Thank you for taking our questions today. It's good to see some progress here. Starting bigger picture, you know, across the quantum landscape, I think peers are pretty divergent on commercialization paths when it comes to quantum computing specifically. Some focus primarily on selling on-prem systems into national labs. Some focus on monetizing platform access through the cloud and application partnerships. With systems now operational in the UK and Japan, alongside the London Forum, can you elaborate on Inflections' commercial scaling model for computing over the next few years. Specifically, I'm curious on how you see the mix evolving between the full stack system deliveries versus maybe cloud and software access.
That sounds good. Yeah, I'll take a crack. And for now, feel free to chime in or Alon as well, or Marcus. And thanks for the question, Jesse. So I'll say this, maybe I'll couch this with, we have to have a little bit of humility when we predict the future here. If you would have asked me this question five years ago, I would have told you I would have thought that in 2026, the vast majority of commercial activity in quantum would be taking place over the cloud or in the cloud. And in reality, it's been the exact opposite. The market's almost entirely driven by system sales. So with that statement out of the way, I think we'll see system sales be a big part of it. For a while, with cloud increasingly coming up as a percentage of the mix. And we'll be, you know, ready to service both ends of the market. But my hunch is early on, customers are going to want these things on their premise. And then they'll, you know, start to shift more to a cloud consumption model over time. And we have customers doing both. But that's kind of my mental model right now. Pranav, what would you add?
The only thing I'll add is just in the context of this week, we are at the Supercomputing Conference in, happens to be in Chicago this year. And tomorrow, I'll be on a panel for cloud deployments of quantum computing. That panel is alongside IBM and Amazon Web Services, which have been leaders in the cloud space. So I think the interest in our hardware is certainly growing quite a bit, and our position on that panel is a reflection of how much cloud interest we've had. But also that event is where the biggest buyers of quantum computing and computing more broadly convene. So we've also brought a display version of our scale QPU that we originally built for GTCDC to that forum. And I think both will see the ground level view up front, but we expect to see high interest in both cloud and system side, system delivery side.
Thanks. I appreciate the color there. And then maybe just as a quick follow-up, and this is actually really quick. I think I noticed you mentioned that there's a growth in backlog in RPO in the release. I think last quarter was 21 million. And I don't know, did you guys disclose what the number was this quarter?
Yes, it will be now 10Q and 51.2 overall RPO No, you'll see now that was five minutes ago.
Oh, that's a great number. Yeah. Can you disclose just between sensing and computing what maybe was the main contributor there? I know it's tough to drill too far down.
I can, you know, there's been growth in both, Jesse, but I think what we were hinting at last quarter, and I kind of said in my prepared remarks, is that Compute has been particularly strong and will be, and we kind of expect to see good growth in RPO in this current quarter as well. I'm feeling good about the way things are shaping up for Compute in 2027.
Yeah, yeah. Frankly, with that number, so am I. Thanks a lot for the time, guys.
Operator
Your next question comes from Craig Ellis with B Reilly Securities. Please see your question.
Yeah, thanks for taking the questions, team, and congratulations on the technical and commercial progress. I wanted to follow up on one of the real notable milestones with respect to today's investor presentation and comments. We're getting to 30 logical qubits very efficiently, and our ambition is to increase that about 3x in 2028 to 100 logical qubits. Can you just talk a little bit about what you've learned getting to the 30 that gives you confidence and that are the leverage points in getting to the 100, and what's the incremental engineering work that needs to be done to get there?
Yeah, sure. In many ways, getting to 30 logical qubits validated the key building blocks that we needed to show for 100 logical qubits as well. So most notably, one of the pieces under the hood is the reconfigurability of our qubits. And that's dynamically, as we run our computation, literally picking up the qubits and moving them around. And that is the core piece of why we have such an efficient ratio of physical qubits to logical qubits. And that's... Efficiency is going to be key to also how we get to 100 logical qubits, with not tens of thousands of physical qubits, but just ones of thousands of physical qubits. The other thing that was validated at this scale was maintaining good performance of our underlying physical gates, and one thing that I drilled into just earlier was on the retention of atoms, or inversely, the loss of atoms. One of the things that we intentionally do is we convert our errors into loss when possible, and this is a really advantageous way to convert ambiguity in our errors into definitiveness of our errors. So this is getting a little bit in the weeds, but the point is that these were key building blocks that were de-risked. Now, getting to 100 lossable qubits. The highest level involves two things. One is we're going to be running much bigger computations in terms of the number of steps at 100 module qubits. And so we need to pick error correction codes with an even greater distance. We showed distance four results. And we need to take that up just a couple of notches. But the key building blocks were already shown at distance four. And then the other piece is that we need to make sure that as those atoms fall out of the system, That is, they are lost. We reload them back in. And this is continued internal progress on the milestone that we showed about four or six months back with atom reloading at constant magnetic field. Technical details are not particularly important here so much as the fact that this is a capability that lets us sustain very long computations. And so, a lot of great work happening that technical teams are performing, and then even greater systems engineering work to bring it together into one demonstration, and that's why we feel great going into 2027 towards our 2028 milestone of 100 logical qubits.
Thanks for getting into those weeds. Very helpful. The follow-up question, really, Matt, is for you. In your prepared comments, And already you've talked about 2027 being a bigger year for compute, and you've got a really strong RPO number. I don't think you're disclosing how much of that is compute. But the question is this. There's obviously a lot that's happening as you talk with potential customers and government entities that are interested. And inflections, neutral item capabilities. So can you just put some color around how that number of conversations has evolved over the last three to six months and how you'd expect it to evolve over the next three to six months, any color around geographic or other parameters that would help us get a better picture of where demands coming in would be helpful. Thank you.
Sure, Craig. A couple of thoughts on that. One, I would just say, you know, we've been consistently saying since we came public and even before, we're always going to take our investment decisions very seriously and be very... Be good caretakers of investors' capital. And so we've hinted that we're going to start to invest more in compute in Q4, and we already started doing that in Q3 and ended 2027. And the reason I feel comfortable doing that is because we're seeing all those leading indicators that you just hinted at, increased levels of conversation globally, so not just in the U.S. And I think you should look to the global markets for where some of this strength is coming from, some of it's coming from the U.S. as well. And the types of conversations and the specificity of conversations and the applications with which people are zeroing in on are getting tighter and tighter as we're approaching 100 logical qubits. And so we're seeing just a lot of really good indicators. And I think the most tangible thing is just the increase in RPO. And now we'll see that RPO continue to trend this year, and compute's been a big part of that.
Thanks, guys. Thanks, Greg.
Operator
Your next question comes from Atif Malik with Citi. Please see your question.
Hi, it's Adrienne Colby for Optus Mallet. Thank you for taking the question. I was hoping you could update us on the timing of the delivery of your third quantum computer to the IQMP Park next year. I guess what I'm wondering is how to think about the mix of compute versus sensing as we move into 2027, given the Innovate UK contract and the extension with NASA, just how we should think about that ramping up, you know, from here to 2027.
I can take a crack, and then, Alon, why don't you fill in? So, our plan is to deliver the system to the Illinois Quantum Microelectronics Park by the end of 2027, and so that will be revenue we'll book in 2027, and you're right, Adrian, that we do have the contract extension from NASA and then the UK. So I think for modeling purposes, you can book that $10 million in the IQMP as revenue next year. And I think what you'll see is that will start to take compute up more as a percentage of revenue in 2027 than 2026. Alain, anything you would want to add to that?
No, I think you described it well. We expect to see the, you know, compute portion of the revenue increasing, you know, next year versus this year.
Great. And as a quick follow-up, as we're starting to think about 2027 in more detail, how are you feeling about the scaling inflection? Where do you think you need to be investing to grow the business to meet the demand you're anticipating next year?
One of the areas that we're focusing a lot on right now, in addition to ramping up our compute R&D, is our manufacturing capabilities. So being able to manufacture not just our sensing products, but our computers in greater numbers than we have in the past because we're seeing that type of demand. So to me, that's one of the things I've been the most focused on. And luckily, we're, I believe, ahead of the curve because of our sensing business and our ability to build things in quantities of larger than just a few. But that's something we need to beef up to be ready to sell, particularly computers in big numbers. So that's the answer I would give.
Operator
Thank you. And your next question comes from Ella Telczynski with Morgan Stanley. Please take your question.
Great. Thanks for taking my question. I wanted to follow up on the comment that customer engagements are shifting to co-development. Are you seeing new types of engagements emerge as a result, and how are those relationships evolving?
Great. Ella, thanks for the question. Pranav, do you want to take that one?
Yeah, sure. I'll go back to this notion of neutral atoms as effectively the FPGAs of quantum computing, and so a lot of the recent interest has come out of a lot of the anticipated buyers of quantum computing realizing that the The initial cost estimates they had of what it would take to run something on a quantum computer, if it was X, it turns out it can be X over 100 or 1,000 thanks to the reconfigurability of neutral items. So I would say that's the... Strong is tenure of the conversations in terms of what brings customers to our desk, so to speak, and what reels them in is the existing tooling that we've built with resource estimation to make it possible to understand how much of an advantage you get with neutral items and also what the runtime is with these problems. And then maybe the other dimension that I'll take this answer in is the industry's understanding of where the earliest applications of quantum computing. Live has also refined in the past year and a half or so. So, for instance, it's generally become a well-understood conclusion that some of the earliest applications are going to be in materials, are going to be in designing better batteries that last much longer, designing better solar cells, designing better catalysts. And so we've seen uptick in demand from those sectors in particular, and we've hired, in anticipation of that demand, some of the world's best material scientists in-house. So that we speak the same language as our customers, and that's been helpful for this co-development of applications, too.
Operator
Your next question comes from John McPeak with Rosenblatt Securities. Please state your question.
Thanks. Congrats on the 30 logicals ahead of schedule, guys, in the lower logical ratio. I'm looking forward to the paper. I'm assuming that the two-cubit gate error rate will be in that paper, Pranav. Is that fair?
Yeah, I think it will be in their least calculations that back into it. The only reason it potentially wouldn't be is that at the scale of operation that we're running at, there are so many gates that we run, and this is not pressing the same button many times, so to speak. This is, imagine a factory floor that has thousands of components. And if we had to measure each of those individually, it would be quite painful. And so we find that a more productive way to actually measure the overall success of the application is at that factory level. It's the entire circuit. And the only way that the circuit can run well, the factory can run well, is if each of those gates individually runs well. And so it's actually quite easy to compute backwards from the application-level fidelity to the two-qubit component fidelity. And again, the only reason it wouldn't be there is just the cost of benchmarking each gate. But we have done recent publications where we've released our underlying two-qubit gate fidelity. We've done 99.73% two-qubit user-facing gate fidelity. And just to give you a quick number here, we ran about 1,000 operations, and our net fidelity after 1,000 operations was 21%. So, roughly speaking, that implies a per gate success rate of about 99.85% or so. Now, that is averaging across many different types of gates, 2QB gates, 1QB gates, motion gates, RZ gates, et cetera. But that's kind of what we want to steer towards looking forward, which is the aggregate impact of every single type of gate, rather than just looking at one type of gate, which is 250 gates.
Gotcha. And then just to follow up on the mega-quap machine, you're going to have about a 20 to 1,000 physicals to 100 logicals. To do a million operations, you're going to have to get 10 to the minus, better than 10 to the minus 7, two-cubit error rate, I think. Will these be universal gates at that fidelity?
Yeah, John, great question. This is a universal system with Megaclop. And we've been keenly studying approaches such as something called the star architecture for getting these universal gates. This is something that's become more attractive to us recently than Procedures like magic state cultivation, which I'm sure you're familiar with, point being that we have some baseline approaches that already give universality, and we've found ways to bring the cost of universality even lower. But the key takeaway is, yes, this mega-quap 100-logical cubic machine will be universal quantum computation every gate that you can possibly think of we could do.
And then just, sorry, will it be the same ion chamber that you guys are using now? Or is there some work that's going to be done on the ion chamber? You're just putting more atoms in there. More atoms, I should say. Not ions. Not ions. Adam, James. Sorry, go ahead.
All good. Yeah, this will be the cells that we've been using. We might make a few tweaks to them. To your point, this is one of the definitive advantages that we have because we make our own ultra-high vacuum cells. In fact, a lot of the quantum industry, including the neutral atom quantum industry, buys them from us. So, we have first dibs on the latest and greatest cells. It turns out that the existing cells are already supporting the capacity of qubits that we need. We might make a few tweaks here and there for different wavelengths, different codings that we need, but the fundamentals have already been de-risked in our current architecture for the, we call it the juniper cell.
Great. Thanks, guys. Thanks, John.
Operator
Your next question comes from Tyler Anderson with Craig Hallam Capital Group. Please say your question.
Hi, everyone. This is Tyler Anderson on for Richard Shannon. Thank you for taking my questions. I was wondering for this contract. Hi, guys. I was wondering for the manufacturing increase in scale, you had previously mentioned that you were interested in engaging with contract manufacturers, and I'm just wondering if this is a priority for your 100-logical qubit machine or your 1,000-qubit machine.
Well, First things first, we're looking at potential contract manufacturers for some of our sensing products, and so that would be where we would be looking first to CMs. In terms of the computers, we'll have to make that decision, and that's part of what we're looking to bring that kind of DNA into the company to help us make that decision, And when it makes sense to engage contract manufacturers, if it makes sense to engage them on the 100 logical qubit machine or to wait until further machines. And so I'd say stay tuned. And that's one of the things we're scrubbing internally, but we're looking at them more actively for our sensors at the moment.
Okay, that makes sense. I look forward to hearing about that. And then I have one technical question. So for the PIC work that you guys had done, please correct me if I'm wrong. But from my understanding, this seems to be for the QCCG sensor as well as the quantum computer. And it seems to be as you move things from the vapor cell. And translating that from a vapor of atoms into individually addressing a single atom. Is that the work that's being done here, including the cooling of atoms?
Yeah, I can take a swing at that one. So I think you're referring probably to your collaboration with Honeywell, is that right? Yes, and UCSB, I believe. Yeah, so indeed, there's slightly different requirements for vapors versus ultra-cold atoms. But in fact, there's also radical similarities in the sense that when we're building PICs and more generally integrated photonics laser systems that can survive the launch conditions and pressures, et cetera, for going into space, we're effectively, fundamentally what we're doing is we're... building robustness into the photonic system, into the optic system. And it turns out that robustness is also what helps us have extremely low noise in our lasers for entangling gates, for instance. So the same things that would make it so that a laser would work surviving the launch pressures and working out into space for, in the case of the QGG sensor, that is actually also cold. But imagine instead if it's the clock, which is a warm vapor. Those advantages actually matriculate to quantum computing as well, because if you have a fundamentally stable laser, you also have a very stable light-matter interaction, and that is how we get long-running quantum computation that consistently stays at, let's say, 99.95%, two cubic hits, et cetera.
That is very interesting. I can see how the minimization of vibration there would help. Thank you.
Operator
I appreciate the answer.
Thank you. Yeah, let's talk.
Operator
Your next question comes from Kingsley Crane with Canaccord Genuity. Please state your question.
Great. Thanks for squeezing me in. So I appreciate all the comments on the value of software on their platform. I'm just kind of curious to your thoughts on The acquisition of Classic by Intel, and then maybe if that reinforces the importance of serving as a hardware agnostic layer, and then just, you know, if customer opinion or conversations have changed at all as a result of that change. Thanks.
So as you may know, but in case not, I can refresh. Inflection was originally ColtQuanta, which was a hardware company. And in 2022, ColtQuanta acquired Supertech, which was the company that I co-founded at the University of Chicago as a software company. And so this heritage of combining hardware and software with very deep co-design has been in Inflection's DNA ever since that acquisition happened. And we see the Intel acquisition of Classic as being very validating to that decision. We're lucky to have made that decision four years back and have had four years' progress to build on top of that. And I think that Intel also sees the... Advantage that you get from having a software team that is working really closely with the underlying hardware team, and in Intel's case, they're working on silicon spin qubits. We continue to have business in our quantum software stack from other modalities, like superconducting and Trapped Ion. Our Superstack platform continues to develop for those modalities as well. But overwhelmingly, we see the highest amount of demand in the Neutral Atom platform, which, of course, we develop hardware for ourselves. And so I would expect that the bulk of our attention on the quantum software side goes into this deep co-design towards neutral atom systems that have software that is tailored to neutral atom. But we are certainly encouraged to see other companies recognizing the massive value of the software stack being so close to the hardware stack.
Okay, helpful. And then maybe just to check, like, any change in tenor or timeline on government's awards based on just political cross-currents and the recent election? I know that you reiterated that, I think, to close that commerce transaction in Q4, but just curious. Thanks.
No, Kingsley, the answer, yeah, no, there isn't. It's saying things are moving forward, and we are getting closer and closer to the finish line to getting that deal linked.
Okay. Thanks for the time. Thank you.
Operator
Your next question comes from Quinn Bolton with Needham Company. Please stay your question.
Hey, guys. Thanks for letting me ask the question. I guess really two questions for Pranav on the 30 logical achievement. First, it sounds like if your sort of factory fidelity is 99.85, that the logical error rate you achieved was roughly 10 to the minus 3. Is that correct? And then if you look forward to the mega-quap machine, I assume your logical error rate target would be 10 to the minus 6. Is that sort of the right levels to be thinking about? And then I've got a follow-up.
Yeah, thanks for the question. That's effectively right. That's the way that I think about it. I'll maybe just give one qualifier, which is that analysis does slightly conflate two different things, physical qubit fidelity and logical qubit fidelity. But for practical purposes, the way that I think about it is exactly the way that you just said, which is we ran 1,000 operations, got to 21%, and therefore, if you back out the 0.2 rates to 1 over 1,000 roughly, you get that 99.5%. And then, sorry, remind me, there's a second piece to your question, too.
Oh, just the mega-quap machine would probably require a logical error rate of something
like 10 to the minus 6 to be able to do a million operations. Bingo. And so that is 1000x better than 10 to negative 3. But the beauty of quantum error correction is that it's exponentially powerful in the number of physical qubits that we add. So getting from 10 to negative 3 to 10 to negative 6 is not a 1000x. It's a challenge. It's, well, again, a bit technical. It's kind of polylogarithmic of that number, which just means it's a lot easier than 1,000x, easier to make this next jump. So we really think of this as the important zero-to-one moments happened, and that's a one-to-end moment. Still work to be done, but that's why we're feeling good about 2028 with a million operations, 10 negative success error rates. Excellent. And the second question was just, you talked about the error correcting codes and the 8 to 3 sort of physical to logical qubit. But I guess I'm wondering, when you said the system did about 1,000 operations, are you actually doing 1,000 rounds of live quantum error correction where you're extracting syndrome, doing the analysis, the decode, and then, you know, generating new Gates or operations based on errors.
Is it a live error correcting in that 30 logical achievement?
These results are with a single round of error correction and error detection. And so the next step towards 100 logical qubits and 1 mega quap is to do many rounds of error correction. And in our eyes, a lot of the key billing blocks have been de-risked. In our control system, in our error correction stack internally, it's also an area of our relationships with companies like Cisco, Riverlane, and NVIDIA, where we've de-risked a lot of the key integrations that are helpful here. But the next frontier is indeed when we do those multiple rounds of error correction and error detection instead of just a single round.
Got it. And that's with the 100 Logical in 2028, where you'll have the multiple rounds. Bingo. Yep. Got it. Okay. Thank you very much.
Operator
Yep. Thank you. Thank you. And we have reached the end of the question and answer session. I'll now turn the call back over to Matthew Kinsella for closing remarks.
All right. Well, thank you, everyone. Great questions, as always. We appreciate everyone's interest in inflection. We've got a lot to deliver in the coming months, and so we're going to focus on execution. Have a great evening.
Operator
Thank you. This concludes today's conference. All parties may disconnect.