Xanadu Quantum Technologies Limited (XNDU) Earnings Call Transcript
November 11, 2026
Earnings Call Speaker Segments
Thank you. Hello everyone, and good afternoon. You're listening to the third-quarter conference call for Xanadu. Joining us on the line today are our founder and chief executive officer Dr. Christian Weedbrook, Michael Trzupek, our chief financial officer, and Rafal Janik, chief operating officer; a live webcast of this call is available, and a recording will later be posted on Xanadu's investor relations website. The information discussed here is time-sensitive and holds true only as of the date of this live webcast, November 11, 2026. Forward-looking statements and information based on current expectations may be part of this call. Significant risks and uncertainties apply to those expectations and that information, so actual results could differ materially from the conclusions, forecasts, or projections contained in the forward-looking statements and information. A number of material factors and assumptions went into forming the conclusions and producing the forecasts and projections that appear in the forward-looking statements and information. Those risk factors and assumptions are covered in the periodic filings we make with the SEC and with the relevant Canadian securities regulators, including our latest quarterly and annual reports, and anyone weighing an investment in our securities ought to review them. Management accepts no duty to revise these forward-looking statements and information, except where the law requires it. At this point I'd like to hand the call to Dr. Christian Weedbrook, Xanadu's founder and chief executive officer, who will take you through the progress we've made on hardware, our technical roadmap, and our partnerships. After that, Rafal will walk you through software, and Michael will guide you through the financials—so with that, over to you, Christian.
Thanks, Brett, and good afternoon to all of you. We enjoy starting out with the mission we hold: building quantum computers that prove useful and are accessible to people in every corner of the world. Xanadu came into being in 2016 with the aim of constructing fault-tolerant quantum computers based on photonics, and at present we have released three generations of hardware. We have over 160 PhDs along with more than 150 patents and patent applications. It's a privilege to collaborate with top industry partners and end users. Our strategy rests on the belief that photonics constitutes the correct basis for a scalable quantum computer, and ten years of peer-reviewed research has strengthened that conviction. Our chips compute without needing cooling, can be built using silicon manufacturing methods, and depend on telecom infrastructure that is already compatible. Beyond that, photonics gives us quicker clock rates, more room to maneuver on error correction, and a machine that is modular and able to network from the very start. When it comes to the measures that matter — clock speed, logical overhead, fidelity, connectivity, and running at room temperature — we believe photonics offers edges over rival approaches. Practically speaking, these advantages mean our approach can get to fault tolerance and to quantum computers that work commercially. As we see it, quantum computing represents a big opportunity over the long haul, one that could create value across a wide variety of industries; but to capture it, developing technology by itself is not enough. Working alongside partners and customers across several industries, we're creating practical applications and demonstrating the areas where quantum computing can provide real value. Xanadu is building a full quantum computing stack as a single integrated platform—spanning photonic chips, fabrication, networking, error correction, software, and applications for end users. Because we develop each layer simultaneously, we're able to manage loss across the entire system, bind hardware and software closely together, and tune performance from the chip through to the complete computer. Since our previous earnings call, the Xanadu team has achieved several exciting milestones that have pushed this mission forward, and five of those are worth calling out. Starting with technology, we set what we consider a global benchmark for photon loss in an optical switch — a significant advance on our loss roadmap. Next, we formed hardware alliances with GlobalFoundries, ASML, and Bluefors that will help us scale up manufacturing. On the software side, we introduced Backline together with AMD and held our latest QHACK developer conference, which drew a record turnout. Finally, we're broadening collaboration with industrial and institutional partners: with Mitsubishi Chemical we showed quantum algorithms for simulating advanced semiconductor materials. RBC and Booz Allen are using our software for workforce training, and we're also working with Lockheed Martin. To wrap up, we've added more resources so we can keep delivering, including new backing from the Government of Canada and DARPA. Our R&D capacity is scaling well and we're iterating faster. That puts us on track to have a commercially viable quantum computer sometime between 2029 and 2030. Before I get into the specifics, let me restate how we think about this and what our approach is. To achieve our mission, we take a full-stack approach, combining our proprietary photonic quantum hardware with PennyLane, the modality-agnostic software platform that we own. That path defines our core objective: building a fault-tolerant, large-scale quantum computer. For Xanadu this is the defining opportunity, and it demands our finest people, resources, and attention. Putting revenue first, ahead of meaningful quantum utility, or selling systems too early, would in the end pull talent and resources away from what matters most to us, which is delivering the hardware roadmap; we gauge progress by how well we execute against that roadmap on our key technological milestones, and then by partnerships that speed adoption of future commercialization. At this point those are much more meaningful KPIs than quarterly revenue; that said, we think there may be chances to bring in revenue from products or licensing that rest on components we are already developing, or have already developed, along the way to building a large-scale quantum data center. In our view this is a low-distraction way to grow a customer base, draw those customers into our ecosystem, and expose them to wider offerings, PennyLane included. Returning to hardware, our roadmap translates that focus into numbers. By 2029 we aim for as many as 200 logical qubits with an error rate down to 10 to the minus 8. In the year after that, we aim for as many as 500 logical qubits with error rates down to 10 to the minus 16. Error rates at that level, we believe, could lead the industry and would unlock more use cases than other quantum companies can offer today. And from 2031 onward, our target is 1,000 logical qubits or more. We have a proven record of hitting our technological milestones. Our Aurora system, published in Nature, connected four server racks into a single fully networked quantum computer over optical fiber. X8 and Borealis, our earlier systems, were also published in Nature. In our view, combining room-temperature operation, manufacturability, and networking gives photonics an obvious route to a quantum data center. Achieving those goals comes down to loss, which remains the top engineering hurdle we face. If we can drive loss below a critical level and combine that with error correction, we anticipate logical error rates falling steeply, leveraging the networking we've already demonstrated with Aurora, and that should enable us to scale logical qubits without any fixed limit. This is mostly an engineering matter rather than a physics one. During our technology roadmap event in August, we disclosed our loss roadmap publicly for the first time. It tracks the combined performance of 17 key optical components, expressed as a ratio between our hardware's loss and the loss level fault tolerance requires. A lower number therefore indicates we are getting closer. Before I go further, I'd like to clear up any confusion about the loss roadmap we released in August. The loss chart underlying it was drawn with each year marking that year's beginning, not its end—so our expectation is to reach 10.2 times at the start of 2027, not at the end of 2027. The chart itself hasn't changed; we're simply clarifying the timeline it implies. Everything else in our roadmap was based on that same loss chart and remains unchanged; we still anticipate as many as 200 logical qubits in 2029, as many as 500 the year after, and 1,000 or more by 2031 and beyond. I'm happy to report that we've achieved considerable advancement toward that objective thus far, and we intend to share an update during the first quarter. Since cutting losses is an engineering effort that unfolds iteratively—it doesn't move forward linearly and won't necessarily deliver quantifiable gains each quarter—we intend to disclose our aggregate loss figure on a yearly basis. Allow me to take you through a few elements of that headway, beginning with the optical switch. At 7 millidecibels, or milli-dB, we consider this a global benchmark: the smallest loss ever documented for an electro-optic switch in its category, reduced from the approximately 20 milli-dB figure we shared a year ago. We'd also like to provide updates on losses for two additional critical optical components in our loss roadmap. To start, the on-chip homodyne detector—the component that reads out a quantum algorithm's result—has cut its loss by half compared with a year ago, and it now stands at 77 milli-dB.
Next, the fiber delay line that acts as a quantum memory in our computer has demonstrated over a 3x decrease in loss since the beginning of the year. A straightforward approach underlies these results: running additional wafers. Each wafer run and tape-out — that is, every occasion we send a chip design off for fabrication — represents one more learning cycle. Additional cycles provide additional data, allow us to validate performance over a wider variety of devices, and assist us in converting design enhancements into manufacturing outcomes more quickly. Two principal materials are used for our wafer runs: silicon nitride and thin-film lithium niobate. So far this year, we have doubled our silicon nitride wafer runs and raised our thin-film lithium niobate wafer runs by 15% sequentially. This establishes a positive feedback loop for loss reduction. Additional wafer runs accelerate our learning, and quicker learning propels progress. Our partnerships have been built on that momentum, and this quarter we are highlighting three. Each tackles a meaningful portion of our roadmap: manufacturing scale, waveguide loss, and cooling. First, GlobalFoundries. This multi-year partnership will scale our ultra-low-loss silicon nitride chips and our superconducting nanowire single-photon detectors on GlobalFoundries' 300-millimeter manufacturing line in Malta, New York. These quicker detectors increase our clock rate and shrink the physical footprint and resource requirements of our qubit factory, which forms part of our system that generates our GKP qubits. A high clock rate also enables us to use less optical fiber, which cuts total path length and moves us nearer to the fault-tolerant threshold. Second, ASML, the world's leading supplier of semiconductor lithography equipment. As far as we know, Xanadu is the first quantum computing company to set up a direct technical collaboration with ASML. We anticipate that ASML's expertise in advanced lithography will assist us in lowering waveguide loss, which is the number one cause of optical loss in integrated photonic circuits. Finally, Bluefors, to develop compact modular cryogenics that bring detector cooling within a data center footprint. It is important to note that our architecture needs minimal cryogenics relative to most other approaches to quantum computing, and this work shrinks the size and cost even further; this design is expected to remove the need for large cryo plants and make quantum computers simpler to scale. As a reminder, our qubits, gates, and measurements all function at room temperature. Cooling is needed mainly for qubit generation, which is essentially the initialization stage of our computer. Partnerships are one way we are building capacity externally. The other is in-house. This quarter, we took a major step with Inception, our new 158,000-square-foot advanced photonics research and manufacturing facility in Toronto, backed by Canadian federal funding under Project Optimism. This facility will include advanced cleanrooms, heterogeneous integration, and packaging capabilities, which are critical technologies for both AI and quantum computing. Our existing packaging facility has already cut packaging turnaround times from as long as three months to approximately 8 hours, and Inception is designed to deliver the same order-of-magnitude efficiencies in testing, measurement, and assembly. We expect to start using Inception early next year. Pulling all of this together, those steps provide a manufacturing base for a quantum data center that is commercially viable. Now I'll hand things over to Raj, who will take you through the progress we've made on software and partnerships. Thanks, Christian. Christian just laid out our hardware roadmap, and software is tightly aligned with it; what we're building is the software ecosystem, the developer community, and the enterprise relationships that will drive adoption of our fault-tolerant quantum computers once they become available. Our priority today is growing adoption, establishing our platform as the foundation for quantum application development, and building commercial relationships early. We think this approach puts us in position to capture a meaningful share of the quantum computing market as the technology matures. PennyLane sits at the center of that strategy. It ranks among the most widely used open-source platforms for quantum computing. Per the Unitary Foundation, roughly 30% of all quantum developers used PennyLane over the past year. Across the PennyLane family, which includes Lightning and Catalyst, we're seeing about 1.1 million package installs each month. PennyLane has also been cited in over 2,500 academic papers. Our developer community keeps growing. This year's QHACK, our seventh global quantum coding challenge, drew record participation with over 3,500 participants; those figures show developers are already using software to learn quantum computing, develop algorithms, and explore applications. As quantum computing moves toward commercial utility, we believe this established developer ecosystem will be a significant competitive advantage. As Christian mentioned, this quarter we launched Backline in collaboration with AMD; it extends our software infrastructure for fault-tolerant quantum computing. Backline is an open platform solution that integrates classical CPUs, GPUs, and FPGAs with quantum hardware to support the real-time processing, decoding, and control that fault-tolerant quantum computing requires. Quantum error correction is one of the fundamental challenges in scaling quantum computers; it requires processing large volumes of measurement data, identifying errors, and providing corrective feedback at extremely low latency. With Backline, we demonstrated end-to-end processing round-trips of under 3 microseconds using commercially available AMD hardware. This shows the ability to perform real-time error correction using standard classical computing infrastructure without dependence on proprietary hardware. Backline is designed to be extensible across different quantum computing architectures, which creates opportunities beyond our own hardware platform. Since launch, we've seen significant interest from national laboratories, classical computing manufacturers, and other quantum computing companies; we believe Backline positions Xanadu at an important intersection of quantum and classical computing and creates opportunities for broader industry adoption of our software. Beyond our open-source ecosystem, we're beginning to commercialize our software capabilities and build direct relationships with enterprise customers. Many of these capabilities originated from tools developed for our own research and hardware programs; with one of the industry's largest software and algorithms teams, we have built significant photonic quantum computing algorithms and high-performance computing capabilities. During this quarter, we hit a significant milestone: revenue started coming in from our commercial software products, supported by an enterprise customer base that is modest yet expanding, and we’re already seeing initial interest in a few areas. The first is enterprise training and workforce development. More and more organizations understand they need to develop quantum expertise in-house instead of depending solely on a scarce group of specialized experts. RBC, which is Canada’s biggest bank, chose Xanadu to provide practical quantum computing training for its staff. In a similar way, Zalyn is putting an initial set of its engineers through training on our software platform. Alongside the workforce development work we do with Lockheed Martin, these engagements show that large enterprises are starting to put money into developing quantum capabilities with PennyLane as the base. The second area is quantum application development and hardware readiness. We assist enterprises in seeing where fault-tolerant quantum computing might create commercial value and how particular applications would run on the photonic architecture we plan to build, using our simulation and resource estimation capabilities. Our customers look at quantum algorithms, work out hardware requirements and execution times, and evaluate likely performance against classical computing methods. In this quarter, JSR, a top semiconductor materials company, signed a statement of work covering a feasibility study of these capabilities.
These engagements really matter, since they assist customers in building a business case for quantum computing and at the same time give us a view into which applications and performance requirements will count most once our hardware scales. Third, we keep broadening the commercial software portfolio by taking capabilities we developed in-house, among them AI-assisted research tools and high-performance simulations, to a wider group of customers. Several of these products are already in pilot and early commercial deployment. Our approach is disciplined. We launch products early, rely on customer feedback to confirm demand, and concentrate investment on capabilities showing meaningful commercial potential. Alongside these products, we still see keen interest from enterprises looking to pinpoint and build commercially viable applications for quantum computing. One illustration is our strategic partnership with Mitsubishi Chemical. Backed by the Canadian and Japanese National Innovation Programs, our collaboration shows how quantum algorithms could accurately model light-sensitive materials for extreme ultraviolet (EUV) lithography, a key process in advanced semiconductor manufacturing. We are now moving that effort toward a software pipeline ready for fault-tolerant quantum computing to aid development of next-generation semiconductor materials. Stepping back, our software business is advancing along three complementary paths. We are growing one of the industry's leading open-source quantum developer ecosystems. We are building critical software infrastructure for fault-tolerant quantum computing, including Backline; and we are forging commercial relationships with enterprise customers starting to invest in quantum capabilities and applications. Software revenue is still at an early stage, yet our progress this quarter shows the growing commercial potential of our capabilities. Our aim is not merely to build software products for today's market. It is to establish the ecosystem, technology, and customer relationships that position Xanadu for long-term growth as quantum computing becomes commercially viable. With that, I will hand it over to Michael.
Thanks, Brad. I'll walk through our operating results, the balance sheet, and the funding behind our roadmap. Every figure I mention is in U.S. dollars. For the quarter, revenue came in at $1.3 million, versus $1.5 million in Q2 2026 and $1 million in the same quarter a year earlier. Revenue will move around from quarter to quarter depending on when work gets done under our various contracts and on partnership-related revenue. As Christian said, we don't expect revenue to be a material driver of operating activities near term; our focus stays on delivering against our hardware roadmap. R&D expense came to roughly $25.6 million, up $5.9 million from Q2; the rise is mainly due to strategic investments in wafer spend for upcoming tape-outs, process development work connected to our new foundry collaborations with GlobalFoundries, and growth in engineering headcount, which also pushed compensation-related costs higher. As we noted last quarter, we still expect R&D spend to grow over time as we bring on more engineering talent, expand chip development and packaging, and improve all of our subsystems toward fault tolerance. G&A expense was about $9.2 million, versus $11.1 million in Q2. That sequential drop in G&A was mainly due to lower professional fees during the quarter, largely tied to transaction and regulatory fees. The quarter included roughly $1.1 million of non-recurring costs, compared with $2.3 million in Q2. Net loss was $32.2 million, versus $42.1 million in Q2 and $20.3 million in Q3 2025. The smaller net loss relative to Q2 mainly reflects the absence of the $10.6 million fair value loss on earn-out share liabilities and the $1.9 million loss on derivative instruments tied to our synthetic ATM, both of which were recorded in Q2. The earn-out was settled in Q2. Adjusted EBITDA loss was $26.3 million, versus $21.3 million in Q2 and $15.4 million in Q3 2025. The increase from Q2 came mainly from higher R&D spending to scale chip fabrication, including wafer consumption and headcount, partly offset by lower G&A. Capital expenditures, which we define as purchases of property, plant, and equipment, were approximately $8.4 million, versus $6.1 million in Q2, primarily reflecting investment in specialized equipment to test, refine, and scale our chip development and packaging capabilities, along with significant investments in refining the subsystems that will enable our fault-tolerant quantum computer. We also started incurring initial costs tied to developing our new Inception facility; we finished the quarter with 361 employees, up from 306 at the end of Q2 and 277 at the end of Q1. We closed the quarter with $286.5 million in cash, down $26.3 million from Q2. We didn't draw on our synthetic ATM facility during the quarter; we raised $67.2 million in Q2, leaving about $232.8 million available for use today. We also bolstered our funding position with non-dilutive capital. We have signed a definitive agreement with the Government of Canada for roughly $140 million in federal funding, and as Christian described, that money funds Inception. Rather than arriving upfront as cash, these funds should come in over time as qualifying R&D investments are made. We also announced provisional approval for as much as $10 million in additional project funding through DARPA's QBI. That is a 1-year contract extension building on our Stage B work, and we expect to put the majority of the funding toward speeding up critical components of our R&D roadmap. Going forward, we expect cost and capital expenditures to rise in Q4, driven by more spend with GlobalFoundries, ongoing hiring, and equipment purchases. We expect headcount to keep growing across hardware, software, and G&A. With that, I'm glad to hand it back to Christian.
Thanks, Michael. We made solid progress this quarter. I believe the operator will now open the line for questions. Please go ahead.
Hi there, thanks for taking my question. I'd like to ask about what you're doing in the AI data center space; so with HyperLight, you said the ultra-low-loss switching platform could offer certain power efficiency benefits; and in AI data centers, we've also noticed another quantum photonics provider sort of taking a similar route. So we'd really like to hear more about that and just what the near-term opportunity might be, plus any customer relationships. Yes, thanks.
Yes, absolutely. This isn't something confined to just this particular piece of IP around the thin-film lithium niobate switch, but I'd say more broadly that one of the unique opportunities Xanadu has is that we're sitting on a substantial amount of photonic IP built over the last 10 years that applies directly in datacom, telecom on the classical side, and then also on the quantum side in quantum communications and sensing; so we are definitely looking into each and every one of these opportunities as they come up, both for existing partner companies and then potential other customers too. But it's a delicate balance, making sure we don't get distracted from the overall technical roadmap; for us, ultimately, we want to build that utility-scale quantum computer that can drive a huge amount of value and societal change. If the opportunity comes up that one of these devices would be commercially viable without distraction from our goal, we would definitely go after it.
Okay, that helps, thank you. Now on the GlobalFoundries collaboration — I believe you mentioned that you'd moved to nanowire single-photon detectors, and as I recall the GKP result published in Nature relied on transition-edge sensors. Does that represent a change? An evolution? Are the two mutually exclusive, or would I simply like to hear more detail on how the platform has developed? Thank you.
Yes, thanks. Good observation on your part. Broadly speaking, either type of device — TESs or SNSPDs — functions as a PNR, meaning a photon-number-resolving detector, and that is precisely what our computer requires. So we had been using TESs, and after that we chose to upgrade them. Essentially, you can view this as a next-gen step forward, and at this point we're relying on SNSPDs. The motivation behind that choice is that it genuinely represents a significant enhancement over the blueprint we had before. In essence, what it gives you now is a much quicker computer. That's the key point. When compared with TESs, SNSPDs operate orders of magnitude faster. Beyond that, that speed permits a reduction in the quantum computer's footprint too. So yes, nice observation. That is precisely what we're doing. Thanks. Let's take the next question.
Thanks, guys. Congratulations on finishing another quarter, Christian and team. I have two questions. The first is about the switch you're developing with HyperLight. How does that — and I'm no expert in optical datacom or telecom — how does it stack up against what's out there in the market? You brought up 7 milli-dB. How much of an improvement is that over products you can buy commercially?
Yeah, it's actually superior in two separate respects. One of those is certainly the loss. So the loss here is significantly lower. Typically, the loss levels that standard optical circuit switches can tolerate are considerably higher than this. What that enables you to do is put a far greater number of channels into a switch. That matters to us given how we choose the qubits we want inside our quantum computer; naturally we want to retain as much of that light as we can, but we also need a certain number of channels coming in so we can select among them. The classical datacom folks can put up with much more loss, though naturally they're eager to move beyond just 16 or 128 channels toward the largest channel count they can get to drive their workloads. The other piece that may have been overlooked is that beyond being much lower loss, it's also considerably faster; so when you look at rival technologies such as MEMS, liquid-crystal switches, or even a few of the integrated silicon photonic switches announced perhaps recently, those run at reconfiguration times measured in milliseconds. Ours run at microseconds or less. That's incredibly important for our quantum computer, where we operate at megahertz to hundreds of megahertz. We have to reconfigure quickly in order to select the right qubits. But naturally, this has a major effect on classical workloads too.
What's your sense of the opportunity in the market for this device?
Well, inside computing it's certainly pretty big; for quantum computing, this is an essential piece of our device. On the overall market size for optical switching, I personally would rather not say, though naturally it's something the team watches closely to see whether, again, there's a chance for us to commercialize a range of different products built on this technology.
Thanks. Hey, yesterday's press release with HyperLight was really nice. You guys have already discussed it quite a bit. What I'm looking for, though, is context that lets us frame up the impact on the roadmap. And if you could specify how many optical switches Xanadu's quantum computer contains, that would be really helpful.
Yeah, we don't disclose that particular figure publicly. What I can say is that the bigger picture is we have a lot of them. And as you know, roughly 17 key optical elements exist where loss propagates through, and overall we need to bring the loss down. As we mentioned in our talk, updates will be given yearly, and I can say that major achievements like this one that push the loss down to something we believe is a world first help us reach and remain on our roadmap. So that's roughly how we'd like to address that question.
All right, I've got that. Then regarding the non-GAAP operating expense run rate for the September quarter, it actually came in substantially below the level we thought you would ramp to. And it does sound like you're going to keep ramping. I'd just like to get a sense of, you know, what operating expense run rate do you expect to be at once you reach the R&D scale you're planning for?
Yes, thanks, Nehal. Good question. You know, we continue not to provide forward-looking guidance. I'll simply reiterate what I said during my opening comments, which is that we expect it to grow during Q4, and you will see those results next quarter.
Thank you so much, and good afternoon. I'd like to pivot now toward the software side. I was hoping to hear your perspective on what Intel's acquisition of Classiq means competitively, and then, tied to that, how do you expect the different segments of the quantum software stack to develop over time? In other words, do you think there's more or less value in particular parts of the stack relative to others?
Yes, that's a fantastic question. What I'd say is that the Classiq acquisition stands as a great proof point for what we've argued from the start, namely that the software portion accounts for a very large share of the problem. We're also really happy that PennyLane remains the top platform, with over 30% of all quantum developers choosing to build on it. That stack breaks into three separate parts. The first one is the user space — the place where developers actually get their work done, the software development kit for quantum applications. That's the piece we refer to as PennyLane. Within it you'll find a wide range of tooling that helps you build algorithms faster, test them more quickly, and do so with better accuracy; we've now been developing this for more than eight years, and we genuinely do hold that leading position. Just as important is the middleware. We frequently refer to it as compilation. If you've been tracking our stack, that's the project called Catalyst. Once more, we group this under the umbrella label of PennyLane, the brand name we use. Still, this is a genuinely thrilling research domain at the moment, since we keep discovering tenfold or better gains in efficiency—both for how long things take to run and for how large a quantum computer you must have—by optimizing how a developer's code is compiled into whatever ultimately executes on the hardware. Then the final piece is the one we moved into most recently with Backline: the core infrastructure that binds together every subsystem required to actually operate a quantum computer. A quantum computer is not merely a QPU standing alone—it is a set of classical and quantum compute resources operating together. Those resources need assorted latencies, extending all the way down to ultra-low latency, in order to enable quantum error correction. And in practical terms, we have by now assembled the highest-performing platform for that purpose, with Backline. Backline matters greatly to our effort to build our quantum computer, though naturally, as quantum computers grow increasingly accessible, we will observe an ever larger share of that value migrating into the software layer; and once again we are delighted to possess that leading framework, which lets us capture it broadly throughout the whole ecosystem.
All right, thank you for that. And then just second, regarding financing in this environment, you didn't issue any shares under the ATM this quarter. Given that, how should we think about near-term funding requirements?
Well, as we've mentioned on earlier calls, we're always watching market conditions and valuation, and we assess it pretty much day by day.
Good evening, everyone. I'd like to step back for a moment from all those little revenue threads you've been describing during this conversation — optical switches, software monetization, and so on — and then you sort of walk it back by telling us you don't want any of that turning into a distraction. So where does that leave us? Should we be concluding that revenue will probably flow from those areas before a commercial quantum computer shows up in 2029 to 2030? Or are you really just trying to point out how much significant IP sits inside the company? I'm trying to figure out what you're getting at when you surface this IP, but then say you won't let it distract you too much. Could you expand on that a little for us?
Sure — I'd describe it as staying the same, in the sense that whatever we develop, we patent it. We're keen to ensure that we hold ownership of it. PennyLane keeps getting developed by us; and those two really are strong illustrations of work that we'd carry on doing whether or not we make money from it; and they matter enormously to the end goal we have of a large-scale quantum computer, which we refer to as a quantum data center; so chances to do things such as monetizing PennyLane might or might not present themselves. As we've said previously, we're certainly open to that. During our first seven or eight years, the focus was entirely on developing the ecosystem surrounding PennyLane and ensuring that adoption barriers were minimal or nonexistent, and we achieved considerable success in that effort. Now that the ecosystem has been established, we have the option to consider ways of monetizing it, and that is something we will evaluate. The same applies to licensing and the intellectual property tied to any of our hardware; the key ingredient behind our success is really the extreme loss-reduction aspect, as we discussed regarding certain optical components on this earnings call. Every one of those areas could potentially be turned into a product, especially since we have to pursue that work regardless for the sake of low loss—or if a customer approaches us enthusiastic about how minimal the optical loss is, we would welcome such conversations. What we do not support—and this may address your question from another angle, Todd—is expanding into quantum security, quantum sensing, quantum internet, or similar fields. Those are excellent areas, in our view, though I believe folks occasionally overestimate how challenging it is simply to concentrate on a single objective and create something humanity has never created technologically — namely, a quantum computer at large scale. So that keeps us fully occupied, and if a business such as Xanadu manages to succeed at it, that would represent the largest share of the market and the largest market tenant by far.
Great. Then regarding the technical updates, you had a few last quarter, and you had some significant ones recently involving optical loss. Just give us a sense of what impact that has on DARPA and of the timing decision relative to Xanadu? Thanks.
Yes, sure. DARPA obviously likes hearing about these results and is aware of them. They also understand that loss is the biggest challenge. I would say that is one part of DARPA. DARPA has other criteria that a company like Xanadu and anyone else needs to meet as well. As they have publicly stated, they want to keep an eye on the technology and how it is progressing; so we are progressing very well. We are in very close contact with them. I believe the end of the program is around November, December, but it could always be a little sooner, as we saw four companies go into the next stage, or it could be later as well; and a lot of that is built on timing, it is built on DARPA's availability, and so we still have things that we need to report to DARPA. So no new update on that, but we will definitely keep you posted if we hear anything.
Hello there, good evening, and thanks for the question. Your algorithms team has just worked out a way to cut the Toffoli count for QRAM dramatically, which is impressive. So I'd like to know: has anything your algorithms team found ever sent your hardware engineers back to alter what they were building? For instance, what's the strongest example of that taking place at Xanadu?
That's a great question. First, on the QRAM — several discoveries have come along that path. One of those was clearly developer-driven. Another one I especially want to call out came out of an agentic AI discovery. We managed to accomplish in two weeks something that classically would have required more than six months of manual labor on our part. That QRAM step forward was a strong proof point of just how fast this whole field is moving and how the tooling we've built in-house has helped drive it ahead. On the other part of your question — the co-design between our hardware architecture and our algorithm development — that is a highly synergistic process. Something that may not be visible from outside: beyond all the great IP we've created on algos and hardware, a genuine strength of this company is the digital twin we've managed to construct of our system. So when the algorithms side is developing a new approach, they can push that down to see how it will affect not only the underlying architecture and the resources needed to run that algo, but also the error-correcting algorithm that will be implemented. Likewise, when our error-correction experts devise new approaches and new logical architectures, we can propagate that into resource estimates for our top algorithms very quickly and clearly. So I'd describe it as a highly synergistic activity across those groups, and we keep advancing the tooling to bind them ever more tightly so we can iterate around that loop faster.
Thanks, that's helpful. You said packaging turnaround dropped a lot. So I wanted to ask: once Inception comes online next year, which part of the process do you think will speed up the most, and what does that mean for how quickly you can move on loss?
Packaging will keep accelerating for sure, but what's perhaps more exciting than the turnaround time alone is the volume we'll be capable of. To reach our next technical milestones that were covered in our analyst-day presentation, volumes will need to rise, and the facility here at our current headquarters simply won't be enough to support that. Another area where we're certainly adding capacity by a huge amount is test and measure. The quantity of chips we're putting through each quarter keeps growing. We have tens of terabytes of data on chips that have been designed, simulated, manufactured, and tested. All of that can feed back into our agentic design and expert design systems to produce the next round of manufacturing runs, but everything depends on having that test and measure capacity. And last, the third major element of the Inception facility is a newer push into heterogeneous integration. That's a capability that simply doesn't exist at this quality level anywhere in the world that we can access. It's the ability to take silicon nitride, thin-film lithium niobate, and our III-V semiconductor-based photodiodes and integrate them into one package, which will be needed to build out both our QPU and the switches a quantum computer requires. Thanks, that's great.
Thanks very much, and I appreciate all of you taking the time to join us today — we’re genuinely grateful. Here’s a quick recap of today: over the quarter, we achieved meaningful progress throughout our business; on the technology side, we set what we believe is a world record for photon loss in an optical switch, which ultimately moves our loss roadmap forward; we also formed important hardware partnerships with GlobalFoundries, ASML, and Bluefors to back our manufacturing ambitions. On the software side, we launched Backline with AMD and hosted a record-setting QHACK developer conference. We keep deepening ties with industrial and institutional partners, among them Mitsubishi Chemical, RBC, Booz Allen, and Lockheed Martin, which shows the expanding range of real-world applications for our technology. Lastly, we bolstered our financial resources with additional support from the Government of Canada and DARPA, giving us further backing to execute against our roadmap. Thanks again — I appreciate it.
That wraps up today's conference. We appreciate everyone taking part. You're free to disconnect at this point. This live transcript is produced automatically, with no human involvement or review.
Read the full transcript via the API
You're viewing the first half of this call. Get the complete Xanadu Quantum Technologies Limited transcript - plus 132,000+ transcripts from 7,000+ companies, speaker segments and full-text search - through the EarningsAPI REST API or hosted MCP server.
Get an API key View API docs →For developers and AI pipelines
Programmatic access to Xanadu Quantum Technologies Limited earnings transcripts and 132,000+ others is available through the
EarningsAPI REST API and the hosted MCP server.
Quarterly plans from $162 - full transcripts, speaker segments, full-text search,
and the /api/v1/transcripts/recent polling endpoint for ETL pipelines.