Good day, and thank you for standing by. Welcome to the IQM Quantum Computers Q3 2026 Financial Results Conference Call. At this time, all participants are in a listen-only mode. After the speaker's presentation, there will be a question-and-answer session. To ask a question during this session, you will need to press star, one on your telephone. You will then hear an automated message advising your hand is raised. If you joined using device audio, click the raise hand button on your screen. To withdraw your question, please press star, one again, or click the lower hand button. Please be advised that today's conference is being recorded. I would now like to hand the conference over to your speaker today, Chiara Mantegani. Please go ahead.
Good morning, and welcome to IQM Quantum Computers Q3 2026 Earnings Call. Joining me today are Dr. Jan Goetz, our Chief Executive Officer and Co-founder, and Jan Kuerschner, our Chief Financial Officer. Before we begin, I would like to remind everyone that today's discussion includes forward-looking statements, such as comments on our product and technology roadmap, customer demand, system deliveries and customer acceptance, bookings, revenue timing, operating plans, liquidity, and outlook. These statements are subject to risks and uncertainties that could cause actual results to differ materially from our current expectations. Please refer to today's earnings release, our filings with the U.S. Securities and Exchange Commission, our listings prospectus, and our stock exchange releases for more information on those risks. We will also discuss certain non-IFRS measures. Reconciliations to the most directly comparable IFRS measures are provided in our earnings release, available at investors.iqm.tech.
With that, I'll turn the call over to Jan Goetz.
Thank you, Chiara, and thank you all for joining us. Q3 was a quarter of strong execution for IQM, supporting our straightforward thesis. We put working quantum computers directly into the hands of our customers and jointly co-develop our product further. Because we optimize and deliver the full stack, we believe this is the fastest path to fault-tolerant quantum computing. This quarter, our thesis continued to turn into contracted deployments and repeatable delivery. During the quarter, we expanded into new geographies, we signed one of our largest contracts to date, and continued to advance our technology roadmap. This momentum reflects a market that is moving in our direction. Let me first comment on the quantum market more broadly. The quantum computing market is moving from research into early commercial deployment.
Our latest State of Quantum report shows a six-fold contract market surge since 2021, with deals reaching $2.6 billion in value. 46% of enterprise buyers are now mandating on-premises architecture. That is exactly the market IQM was built for. Since 2021, we have been deploying full-stack quantum computers directly into research, HPC, and AI data centers. We have done and are working on 10 quantum AI integrations, where we connect our quantum computers directly to GPU clusters. We do this in supercomputing environments, both with NVIDIA and AMD systems. These integrations are a market-leading number of quantum AI integrations, which is relevant for the current development around AI factories. Against this backdrop, let me turn to how we performed. We delivered against the three main drivers we presented in our August earnings call. First, commercial traction. Second, product development. And third, financial strength.
My remarks will focus on what advanced in each of these three core areas during the quarter. Let me begin with the headlines. First, commercial traction. In Q3, we achieved EUR 42 million in new order intake, adding to an already strong order backlog. Out of the backlog, EUR 9 million was converted into revenue for the quarter. This means a net growth of EUR 33 million in backlog for Q3. When it comes to the actual customers behind these figures, we booked the first order for IQM Halocene, our error correction platform. The order came from CSC and the LUMI AI Factory, a contract roughly equal in value to our entire 2025 revenue. We also announced a new system order in Japan from TOYO, a repeat enterprise customer. In addition, we expanded into a new geography, Brazil, which extends IQM's presence to major customers across four continents.
Second, product development. We produced 250-qubit customer systems, which left our factory and are now being installed at the customer sites. This is a key signal to the market about our ability to deliver on our way to error correction. This ability to execute is underpinned by our ongoing fab expansion, which is on track to produce our next-generation processor generations. And third, financial strength. We closed the quarter with EUR 260 million in cash and have advanced with several grants, leading to additional source of income on top of our sales. Our CFO, Jan Kuerschner, will take you through the numbers. Let's look more closely at our commercial traction, how we drive quantum adoption through use case development, and grow sales through the most important order of the quarter. Customers use our quantum computers to get quantum-ready with their organizations.
They co-develop use cases, they run case studies, and integrate our quantum computers into their workflows. On the enterprise side, adoption is building through skilled and industry applications. Our IQM Resonance Cloud now has more than 17,000 registered users, and it is increasingly where enterprises build their first quantum capabilities. For example, Intesa Sanpaolo, Italy's largest bank, launched an advanced quantum algorithms program through IQM Resonance Cloud that runs on IQM Aurora, our 54-qubit system. And in July, together with Deutsche Bahn, we published a hybrid quantum classical pipeline for rolling stock planning on real operational data, running end-to-end on IQM hardware. Let's put this into context. At these problem sizes, the performance of quantum computers matches the performance of conventional computers. We are building trust by showing that the technology works for subsets of the actual problem size.
It is the working pipeline, the trained team, and the well-framed problem that delivers value to our customers. This is how we will continue to talk about applications, openly and honestly, while pushing the technology further. What enables these use cases are our state-of-the-art quantum computers, which we deploy globally. A good reference point is that our customers and partners have published in total over 100 research papers on IQM quantum computers that we have deployed globally. In addition to quantum adoption for use cases, we drive it by deploying error-corrected quantum computers to customers. In July, the CSC Supercomputing Center in Finland selected IQM to deliver an IQM Halocene system. It is named LUMI-IQ, integrated into their LUMI AI Factory. LUMI-IQ is jointly owned by Finland, Czechia, Norway, and Poland, and co-funded by the EuroHPC Joint Undertaking.
It will be hosted in CSC New Data Center in Kajaani, alongside one of the world's most powerful supercomputers. The total value of this order was EUR 33 million. To make this important deal possible, we announced in September an upgrade path for our quantum computers. LUMI-IQ will be delivered in three stages. First, an IQM Halocene H4 system in 2027 with 150 physical qubits and early error correction capabilities. Then, an upgrade in 2028 to lower logical error rates and enable real-time error correction. Finally, an IQM Halocene H5 system in 2029. After the final stage, the system is designed to support logical qubits with fast real-time feedback and to run Clifford operations between them. This will enable computations with up to nine fault-tolerant logical qubits using distance three surface codes and other more advanced codes, including color codes. Now, why is this important for investors? Three reasons.
First, it is a multi-year, sovereign-backed customer relationship that grows with our roadmap. Second, it shows how system value steps up as we move from today's NISQ machines to error-corrected ones in the future. A single Halocene contract is worth a multiple of a typical Radiance system. Third, LUMI-IQ is on track to become Europe's first superconducting quantum computer with logical qubits, built, owned, and operated in Europe, all with the support from IQM. Especially the third point, the advanced ownership model, is gaining traction globally. In September, we announced that TOYO Corporation in Japan ordered its second IQM system, an IQM Spark for its R&D center in Tokyo. This system joins the IQM Radiance 20 system that will be installed in Japan's National Institute of Advanced Industrial Science and Technology (AIST) quantum center.
Both systems are planned to be delivered over the next three months, and operations are planned to begin in early 2027. What stands out here is how TOYO will use the IQM quantum computers. It is opening both machines to Japanese universities, startups, research institutes, and companies as a test bed for algorithms and for the components that go into superconducting quantum computers. Japan's national strategy targets 10 million domestic quantum users by 2030. Our approach to deliver shared and owned infrastructure supports this strategic goal. In the U.S., our IQM Radiance 20 system at Oak Ridge National Laboratory, called Pathfinder, is now live and operational. Again, it is owned and operated by Oak Ridge's own team. For the first time, the Quantum HPC software stack that Oak Ridge developed against simulators is running on a physical quantum computer.
We also announced a completely new geographical expansion, our first quantum computer in South America. The Eldorado Research Institute in Campinas, Brazil's largest technology cluster, will install an IQM Spark in the Q1 of 2027. They also receive cloud access through Resonance platform to our 54-qubit system in Europe. Like TOYO, Eldorado will open the system to the universities, startups, and industry partners around it. In summary, this means four customers, four continents, and the same underlying decision. They grow through ownership of a full-stack quantum computer with an open software architecture. This is how IQM expands quantum ecosystems globally and delivers technology readiness. As of today, IQM has cumulatively sold 31 full-stack quantum computers worldwide, including Community of Madrid that was announced after the quarter end in October. We have delivered 17 systems to customers, with additional 17 systems currently in production.
Next, let's talk about sovereign programs and our U.S. footprint. Governments are treating quantum as a sovereign capability. Confirmed public funding worldwide now stands at about $41 billion. Governments are no longer just funding quantum, they are investing in it and buying it, and we see this in our pipeline. In the U.S., we opened our first quantum technology center in the University of Maryland's Discovery District, and we welcomed Governor Wes Moore there. In the same building, Microsoft Quantum has opened a quantum hardware makerspace, and IQM is one of its seven initial partners on a quantum hardware training initiative. We joined alongside AMD, Bluefors, Intel, Fermilab, Riverlane, and Quantum Motion. With efforts like this, we help resolve a key bottleneck in our industry, the number of people who have actually built and operated quantum hardware.
In Germany, the Federal Ministry of Research, Technology, and Space launched its quantum computing competition with up to EUR 640 million in planned funding to develop at least two error-corrected quantum computers for industrial use by 2030. The SuperPilot Consortium, coordinated by IQM together with Infineon Technologies and Zurich Instruments, has been selected as one of six projects to advance to the next phase of the competition. It combines our experience in building and delivering complete systems with Infineon's semiconductor manufacturing expertise and Zurich Instruments' industrial-grade quantum control. On the European level, IQM has been selected into the first phase of the Grand Challenge Program. This is part of a renewed EU strategy on quantum computing, which will be implemented through the EU Quantum Act. Finally, in Japan, we are supported by the government through NEDO, a joint project with Kyocera and AIST.
All these projects support the development of our roadmap, both financially and through joint R&D resources. Now, let me switch gears and turn your attention to our technology roadmap towards fault-tolerant quantum computing. Here, our message is clear and consistent. We measure progress at the level of the system, not a single hero number. In May, our work demonstrating above 99.9% fidelity across all three core operations in a single device was published in the peer-reviewed journal PRX Quantum. These results were also presented by our team during the SQA Conference in Gothenburg in August. SQA is a globally leading tech conference for superconducting quantum computers and algorithms, which we co-hosted together with Chalmers University and SCALINQ. The results we presented, a combined fidelity of 99.8% across all three core operations, made up of 99.98% for single-qubit gates, 99.94% for readout, and 99.93% for the two-qubit CZ gate.
The CZ figure is the one I'd like to underline. It isn't a best single measurement, but an average over 40 hours of continuous operation. The gate ran stably for nearly two days. Only one single-qubit recalibration and stayed within a few hundredths of a percent within its peak. That kind of stability is what data center integration, long hybrid workflows, and real applications depend on, and it's the performance we're engineering into our next-generation customer systems. Based on the fast innovation cycles enabled by our own fab, we have reached record T1 times of 1.6 milliseconds. This is a major stepping stone towards higher gate fidelities in our processors. We know investors are focused on how this performance carries over to larger processors. A good example is that we have now entangled all 54 qubits on an IQM Emerald processor after fully entangling our 20-qubit Garnet system earlier.
The new 54-qubit result has been achieved with advanced error mitigation that came from bringing our optimized control methods to full processor scale, together with firmware upgrades and cryogenic refinements. Because Emerald is available through our IQM Resonance Cloud and through Amazon Braket, this isn't a lab result. It is a production system with significantly increased performance to run algorithms researchers and customers around the world can access today. Our priority for the coming quarters is to demonstrate these results also at the system level on our 150-qubit processors. The first two systems are on schedule for final delivery by year-end. We have started unboxing these systems at the customer site and are preparing them for final installation. The next step on our tech roadmap is implementing error correction for our Halocene product line.
We made significant progress on the two things that matter as processors scale towards error correction. The first point is running many gates simultaneously without them interfering. This means we have been suppressing leakage and crosstalk significantly. The second point is connectivity. We have achieved a new direct two-qubit operation between any pair of qubits on our Star topology. Through its central coupling element, the Star topology is the building block of our broader Constellation architecture, where each qubit connects to 12 other qubits rather than to just four qubits. This increased connectivity is the basis for the barbell error correction codes we announced in June. Error correction is where the industry attention has moved, and we are executing a strong roadmap towards fault-tolerant quantum computing. This quarter, our work has focused on making it real-time, open, and testable.
Real-time because IQM Halocene runs NVIDIA Ising Calibration, which we contributed and developed together with NVIDIA natively through NVQLink. This way, calibration and error correction loops run on hardware the customer owns. It is open because with NVIDIA CUDA-Q, which we adopted in September, the same workload can be compared across codes and architectures on common metrics. It is testable because researchers can now run these experiments themselves. Our Qrisp package carries the same code from simulator to IQM hardware. At NERSC Supercomputing Center in the U.S., researchers ran their own surface code memory experiments on IQM Resonance. Based on these achievements, we continue executing our ambitious technology roadmap. In 2027, we target our first error correction demonstrators, followed by first customer deliveries, typically 12- 18 months after each demonstration.
Deals like the LUMI-IQ system for CSC show how that roadmap now translates into contracted customer deployments. Another key ingredient of building fault-tolerant quantum computers for real use cases is the software stack. It is the applications running on top, and of course, AI. IQM's acquisition of Quantistry's assets in July added a cloud-native, AI-powered chemistry and materials platform and an experienced team. We are now integrating that platform into our product solutions. It lets chemists and engineers, not just quantum specialists, run their own use cases, with each step routed to classical HPC or quantum as appropriate. On stable IQM hardware, that is how our customers will identify the use cases where quantum genuinely makes a difference. Our application teams delivered results this quarter that show how hybrid quantum-classical workflows are heading. Let me give you two examples how we drive use case adoption with customers and partners.
First, with Volkswagen Group Innovation, we cut the classical post-processing of a quantum-classical chemistry method with an estimated 248-fold runtime speedup for 100 orbital systems. We validated the results on IQM hardware and applied it in 40-qubit simulations of a reaction pathway relevant to lithium-air batteries. Second, already some time ago, the German Aerospace Center and HQS Quantum Simulations simulated magnetic excitations in real two-dimensional materials on our cloud at up to 48 qubits. Now, this quarter, we delivered the other half, a classical reference pipeline for strongly correlated materials where standard methods break down, giving a rigorous baseline any future quantum materials results will have to beat. These use cases are enabled by several new methods our teams have developed. Let me give you three examples for that. First, our new QISS method extracts better optimization solutions from today's processors already produce.
On our 54-qubit Emerald processor, it delivered the same solution quality from noisy hardware as from noiseless simulation. Simulations of deeper noiseless circuits show how it can be competitive with established classical solvers. Second, we released open-source error reduction tools for IQM systems that reduce errors without amplifying noise, with one line of code and fully reproducible results. Third, we introduced quantum channel polynomial processing, a new algorithmic framework. This framework lets users trade circuit depth against repetitions, so the same algorithms can scale from today's machines to fault-tolerant ones. In its most precise setting, circuit cost grows only logarithmically with the target accuracy.
The common thread here is a view our Chief Scientist, Inés de Vega, set out well this quarter when she said, "Currently, the quantum computer belongs inside a workflow that remains largely classical, alongside CPUs and GPUs in AI and HPC infrastructure." That is exactly where our systems are deployed. We continue developing an open and modular software stack in co-development with our customers and partners. The open-source project Qrisp that we are supporting is growing and has won the Quantum Effects Award 2026 for quantum software, a testimonial for the visionary software architecture we are building. All of these efforts I described so far help our ecosystem to grow. The IQM Research Award drew applicants from 10 regions. Our free Quantum Summer School put new users on real hardware, and after quarter end, we signed a skills and research agreement with Swinburne University of Technology in Australia.
Finally, our people are our most valuable resource. I was super proud to see IQM being recognized with two gold and one silver award at this year's Employer Brand Management Awards. We continue building a world-class team in all our global sites to deliver world-leading quantum computers for the well-being of humankind. With that, I will hand over to our CFO, Jan Kuerschner.
Thank you, Jan. Let me talk more about our financials and how we are executing on our revenue projects. As a reminder, we operate in a project-driven market where annual revenue comes from a number of large contracts. We recognize revenue from sales of large customized quantum computer systems over time and recognize revenue from standardized products at a point in time when control transfers to the customer. Quarterly revenue can therefore look much more volatile than the underlying business, and we continue to guide you to look at three metrics together. First, order intake in the period. Second, revenue recognized in the period. Third, order backlog at the end of the period. Let's review the Q3 results. Order intake.
In the Q3, we booked EUR 42.2 million of new orders, led by the LUMI-IQ Halocene contract, as well as systems for Eldorado Research Institute of Brazil and the Community of Madrid in Spain. This brings year-to-date order intake to EUR 52.9 million. Revenue. Q3 revenue was EUR 8.9 million and EUR 17.7 million for the first nine months of 2026, compared with EUR 11.9 million in the first nine months of 2025, meaning year-over-year growth of 49% for the nine months. Order backlog. We ended the quarter with an order backlog of EUR 102.5 million, compared with EUR 69.1 million at the end of June. We expect EUR 59.5 million of our order backlog as at the end of Q3 2026 to be recognized as revenue within the next 12 months.
Revenue from hardware and services was EUR 8.7 million for the quarter and is EUR 17 million year-to-date. The remainder comes from the cloud-based usage of our quantum computers and from co-development projects, all quantum computing related. Gross margin was 47% for the quarter, compared with 46% in the Q2. Operating loss was EUR 171 million for the quarter and EUR 232 million for the first nine months. As we flagged in August, the Q3 includes EUR 13.5 million of one-time transaction costs related to the completion of our listing. Additionally, under IFRS, we incurred a one-time non-cash share-based listing expense of EUR 127 million during the quarter. This represents the non-cash premium of obtaining a public listing via the merger with RAAQ, and it represents the excess fair value of shares issued by IQM in exchange for the net assets of RAAQ.
The underlying growth in our cost base reflects deliberate investment, R&D for our roadmap, to build our sales and delivery capacity, and our first full quarter of public company costs. We ended the quarter with a total workforce of 490 employees and external staff, compared with 460 at the end of June. Let's now turn to the balance sheet. We ended the quarter with cash and cash equivalents of EUR 260 million, compared with EUR 309 million immediately following our listing on July 2nd. The change reflects operating cash outflow of EUR 48.3 million. Q3 operating cash flows include EUR 6.2 million associated with the July listing. Our cash position provides a runway into the Q2 of 2028. We are reiterating our full year 2026 guidance. Full year order intake of EUR 65 million-EUR 75 million and full year revenue of EUR 42 million-EUR 47 million.
As we said in our August earnings call, 2026 is structurally Q4 heavy. The largest part of our full year revenue is tied to the scheduled delivery and customer acceptance of our first 150-qubit systems. We remain confident to deliver the 2026 revenue numbers, given that the two systems have already left our factory and are currently being installed at the customer sites. Because acceptance-based recognition is the main source of timing risk in any quarter, I want to be transparent about how it works. For standardized systems, we recognize revenue at the point when the customer accepts the system, not when it's shipped or installed. Acceptance depends partly on factors on the customer side, such as site readiness, cooling and power infrastructure, and public sector procurement sign-off. Larger customized projects are recognized over time as milestones are reached.
If an acceptance moves by a few weeks across a quarter end, the revenue moves with it. The contract value, the backlog, and the cash profile do not change. As a reminder, revenue only includes commercial sales contracts of quantum computers, cloud, or related services. We are not counting government grants, public funding, or other non-dilutive financing as revenue, just plain commercial contracts. We include grant funding as other income in our financial statements. Let me also provide two structural points that will further help with modeling our business. First, fewer larger contracts. As our systems move to higher qubit counts and embedded error correction, contract values rise. A single Halocene system is worth a multiple of a Radiance system.
That improves the quality of our order backlog, but it also makes order intake and revenue lumpier and extends conversion from order to revenue to sometimes more than two years. Second, a growing recurring base. As our installed base expands, service maintenance and upgrade revenue becomes a larger share of the mix. LUMI-IQ's staged upgrades are a good example of how one customer relationship generates revenue across several years. Our capital allocation priorities are unchanged. Delivering to our customers, achieving our Halocene error correction milestones, expanding fab capacity, and deepening our cloud and software ecosystem. We will remain disciplined and opportunistic M&A that strengthens our offering, as we did with Quantistry. Our focus is on building an enduring, long-term profitable leader in quantum computing infrastructure. With that, I will hand the call back to Jan Goetz.
Thank you, Jan. Let me close with three takeaways from the quarter. First, our commercial and geographic base is broadening. We announced and won new orders in Japan and Brazil. Our first U.S. system is live at Oak Ridge National Laboratory, and enterprises are building skills on IQM hardware through our cloud, while sovereign customers in Europe remain our anchor. Second, our roadmap to fault-tolerant quantum computing is continuing to be contracted. With LUMI-IQ, a customer is not just buying a single machine, but committing to a multi-year path toward logical qubits on IQM hardware, with system value that steps up as performance does. And third, the Q4 is about execution, delivering and achieving acceptance of our first 150-qubit systems, converting backlog into revenue, and building order intake for 2027. For years, the open question in our industry was whether quantum computers could work at all.
Today, the question is whether they can be built, delivered, and operated at the pace the world needs them. That is a manufacturing and deployment question as much as a physics one, and it is exactly the question IQM has spent years building the capability to answer. Thank you all for joining us today. Thank you to our customers and partners, to our teams in Espoo, Munich, College Park, and around the world, and to our shareholders for your trust. Our earnings release and presentation are available at www.investors.iqm.tech. Operator, please open the line for questions.
Thank you. We will now begin the question-and-answer session. As a reminder, if you wish to ask a question, you will need to press star, one on your telephone, or if you have joined via device audio, click the raise hand button and wait for your name to be announced. To withdraw your question, please press star, one again, or click the lower hand button. We will take our first question. Your question comes from Tyler Anderson from Craig-Hallum Capital Group. Please go ahead. Your line is open.
Hi, this is Tyler Anderson on for Richard Shannon. Hello, everybody. It is great to talk to you again. I wanted to characterize the existing deployments that you already have. For the systems that are already in production, how many of these groups are error correction research groups, or at least that was the intention as to why they purchased the system? Then I have a follow-up.
Yeah, thanks, Tyler, for the question. The deal that we announced with LUMI CSC, this is a Halocene deal, which is really targeting the error correction functionality. All the other systems that we have sold are either Radiance or Spark. This means, in principle, customers and users, they can run error correction workload there, but they have to implement everything themselves. So it is really basically their development they have to put in. Whereas these new Halocene systems, this is where we bring in the error correction capabilities.
Okay. Are you also supplying the error correction hardware for this?
You mean for the Halocene machine?
Yes.
Yeah, this is an upgrade path that the customer takes there. As part of the upgrade, also the error correction capabilities will be deployed. So, for example, the real-time part, I think, will only come with the second delivery and so on. The system kind of will grow in functionality over time. But at the end, yes, there will be full functionality in the system with the customer.
Okay, that's great. Then you had mentioned you had open-sourced error reduction tools. Are you speaking to error suppression, or is this error mitigation?
Yeah, so we have basically put these tools also on our cloud, and people can use it. The focus here has been error mitigation. I'm not 100% sure if there's also error suppression in there, and this is something we would need to double-check.
Okay. Thank you. I'll hop back in the queue.
Thank you. We will take our next question. Your question comes from Jesse Sobelson from U.S. Bancorp. Please go ahead. Your line is open.
Hi, guys. Hi, everyone. Thanks for taking my questions today. You reported 31 systems sold, 17 delivered, and 17 are currently in production. I think that implies that there's three that are being produced with no end customer in mind. Are those captive systems that you're building, or what's the reason behind this?
All systems have a customer tied to it. There's no systems that we're just building for the shelf.
Yeah. I guess I'm just curious on, sorry, with the deliveries that you have, can you provide a status update? I know that there was a two-system CESGA contract in Spain. We were expecting those deliveries to come through in June. I got the commentary here on the 150-qubit systems being delivered to customer sites, but is there any detail on the CESGA contract?
Yeah, I think so. Well, first of all, the other two 150-qubit systems that are currently being deployed, so they are already at the customer site. The CESGA system is not yet delivered to the customer. We are facing a slight delay here, but we are not worried about that. They will slip into the beginning of next year, and that does not concern our revenue guidance that we were giving on that.
Okay, great. Yeah, I think that it speaks really well to you guys having the 150-qubit machines on site, and I think it is looking very interesting for the Q4. So, I appreciate you guys taking my questions. Thank you.
Thanks. You are welcome.
Thank you. Your next question comes from the line of Felix Henriksson from Nordea. Please go ahead. Your line is open.
Hi, guys. I have a couple questions. Take them one by one. Firstly, the LUMI upgrades that you mentioned, are those included in the EUR 33 million contract value, or are those incremental revenue on top of that?
Well, thanks for the question, Felix. Yes, those are included in the EUR 33 million.
Got it. That's very clear. Then, you called out the IPO transaction-related costs, EUR 13.5 million for Q3. Are there any sort of one-time, non-recurring listing or other related costs that we still should expect to be included in the Q4 financials? Thank you.
No. By now, we've recorded them all.
Great. Crystal clear. Thanks.
Thank you. We will take our next question. The question comes from Mike Harrison from Rothschild & Co. Please go ahead. Your line is open.
Oh, hi, guys. Thanks so much for taking the question. They are more sort of financially focused than technologically focused. The first is, if I take your ordered intake and revenue guidance, I think it is implying your backlog will fall to about EUR 90 million by the end of the year. So I just wanted to double-check that that tallies with your expectations for what Q4 looks like. Then secondly, if I take what you are saying about the amount of backlog that gets recognized in the next 12 months, I think that is implying that your nine-month 2027 versus nine-month 2026 revenue growth rate would be tracking at about 85% year-on-year.
I just wanted to kind of sanity check with you guys whether that is a roughly sensible glide path for how you think the quantum market evolves, whether you think that is a bit fast or a bit slow. Just kind of helping me understand how I should contextualize that bit.
Okay. Let's start with the first part. As we repeated our guidance, one could make the math and say that approximately EUR 20 million-EUR 25 million of revenue will be recognized in Q4, and that's what we expect. If you deduct that from the EUR 102 million that we have currently as of the end of the quarter, that doesn't arrive at the EUR 90 million. However, that being said, we are still expecting some more orders to come in until the end of the year, so that would then again raise that number. Here, as we gave our guidance, we have the EUR 53 million now secured, and it will be anywhere between EUR 65 million and EUR 75 million. So if you add that to this, your math kind of seems reasonable.
Maybe I can comment on the market question. I think overall, of course, it's difficult to derive how the market that's based on the performance of how we are doing. Overall, of course, we are super focused on execution, and this usually includes two points. One is the product and technology development, so building better computers, and the other one is the commercial traction. But still, there might be fluctuation quarter to quarter, so I think there's some uncertainty here. The other thing is, of course, that we are really pioneering the field of deploying quantum computers to supercomputing center, and doesn't mean that the whole market is kind of following at the same pace necessarily.
Okay, thanks very much. I do have a follow-up, but I'll jump back in the queue.
Thank you. We will take our next question. The question comes from Krish Sankar from TD Cowen. Please go ahead, your line is open.
Hi, yeah, thanks so much for taking my questions. This is Steven calling on behalf of Krish. The first one I have is regarding the new order intake for the quarter. Just kind of curious, besides the IQM Halocene system for LUMI AI and the TOYO, and also the Brazil orders, were there any other system orders beyond that within the EUR 42 million intake? Or is the remaining balance potentially from cloud bookings and future system usage?
First of all, thanks for your question, Steven. There is some cloud in there, but not really material. There are some other systems in there which we have not yet announced, but that will be happening soon.
Got you. Helpful. Then as my follow-up, in terms of error correction milestones and strategy, I understand you guys are pursuing and developing multiple technologies, surface codes, color, and also barbell codes. I guess for the primary development on the IQM Halocene systems, are barbell codes the main focus there, or is there going to be a meaningful amount of resources focused on traditional surface codes? How should we think about milestones next year for development of error correction technologies? Specifically, if you could talk to what kind of, I guess, physical qubit to logical qubit ratios you might be able to achieve by the end of next year. Thank you.
Yeah, well, thanks for these questions. I will take them. On the error correction approach, I think it's always important to check where we are coming from and then where we are planning to go. When we started, the whole error correction research was not as developed as it is today, and kind of the leading approach was the surface code. This is why we designed our first systems to support this, and this is basically the Crystal architecture that we have chosen. All our systems that we have shipped so far, except for the Czech one, are actually targeting this. But then we have published these very nice results on the barbell codes, so a different approach, which is more efficient. The goal is really to lean towards the barbell codes as we move on.
But in principle, of course, you could even think about systems where you have, in a way, both. Because on the barbell codes and the QLDPC in general, there are still some, let's say, topics which are open to research. For example, the magic state distillation. We know that it works with the surface code architecture, and let's say, a de-risking or fallback option would be to do this part still with the surface code, for example. So in a nutshell, the goal is to really move to the barbell codes, but having the surface code in a back pocket to de-risk the approach. When it comes to this ratio between physical and logical qubit, this really depends also on the target error rate on the logical side that you are aiming for.
It is also a moving target, because it also depends heavily on the physical error rate. So it's really hard to give one number that captures it all. Overall, the barbell codes that we have developed are, of course, much more efficient than the surface codes. This means you really kind of gain there order of magnitude in this ratio between surface code and barbell codes.
Thank you so much, Jan.
Thank you. We will take our next question. The next question comes from Waltteri Rossi from Danske Bank. Please go ahead, your line is open.
Hi, guys. Waltteri Rossi from Danske Bank. My question is about cash flow. You state that the current cash runway extends into Q2 2028. What assumptions for revenue, customer advances, and capital expenditures do you have that underpins that runway? A follow-up, looking at your year-to-date operating cash flow, it was roughly EUR 99 million negative. How much of that is recurring operational cash burn, and what should we expect to be this kind of normalized cash burn in Q4 and 2027? Thank you.
Okay. Well, thanks, Waltteri, for your question. We are not yet giving guidance on 2028, since you were asking on revenues, what to expect for that year. It is a bit difficult to answer that. We do, of course, anticipate a growth in revenues, and as from the previous questions and our guidance that we give on the order book and what will be recognized over the next 12 months, we can derive a number here. However, as I said, we are not giving guidance yet, and so that I will need to ask for your patience for our next earnings call here.
Concerning this year and the operating cash flow, we were stating that there are some expenses tied to the listing, and that was EUR 13 million, so the rest of that operating cash flow is the regular cash flow. We can interpolate the same for Q4 in that sense.
All right. That is helpful.
You probably expect something more concrete, so sorry for that, but yeah.
Yeah. Understood. Thank you.
Thank you. Once again, if you wish to ask a question, please press star one one on your telephone or click the raise hand button on your screen. We will take our next question. The next question comes from Tyler Anderson from Craig-Hallum Capital Group. Please go ahead. Your line is open.
Hi, everyone. Thank you for taking my follow-up. I guess what I am really trying to figure out here is we have this Star topology that can now do error correction, and from my perspective, if I was doing error correction research on superconductors, that would seem to be a thing that I would want to get into so that I can develop on the platform for the future. Now, we have this in 2027 being delivered, and then there is the mix of the Crystal architecture and the Star topology, which again, seems like one of those, how should I say, like a change in regime, just a changing tide in how things are done.
What I am really trying to find out is, with this install base that you have, it would seem that there are a good amount of people who would want to upgrade to the new system, or at least one of the two systems. Also, it would seem that this would attract new customers as well. I just want to kind of get an understanding, where do you have your expectations for Halocene, and do you think this is going to attract more interest from customers that you have not reached yet, or do you think that this is going to drive an upgrade cycle with those who already have an installed and up and running quantum computer?
Well, thanks. I think that is a very valid question, and generally, we do think both will happen, and we have seen both happening. We have customers, they went through the upgrade path, and we have good indications that this upgrade path is also attractive for others. We do think that our current customer base will continue to invest into our roadmap, and this is one thing. You have seen the numbers. We sold 31 systems. Some of them are already upgrades, so the number of customers is smaller. If you just count the number of supercomputing centers, HPC centers, data centers around the world, obviously, you come up with a much higher number. There is also, I think, a good potential to open new markets and gain new customer traction, as we have shown this quarter.
With Brazil, for example, we have now the first customer in South America, and we do think that there is a way to do both, to serve our existing customers with new and better products, and at the same time, of course, then win additional customers on top of it.
Thank you for the color, and also thank you for the logical memory experiment. I appreciate it when we point out the memory.
Thank you.
Thank you. There are no further questions. I will now hand the call back to Jan Goetz for closing remarks.
Yeah. Thanks again, everyone, for joining. If you have further questions, I refer you to approach our investor relations team, and they are happy to help. With that, I am really looking forward to connect then to the next earnings call for the next quarter. Thank you.
This concludes today's conference call. Thank you for participating. You may now disconnect.