Quantinuum Is Selling Its 2029 Computer Now, and a Bank Is Buying
QuantumAnalysis

Quantinuum Is Selling Its 2029 Computer Now, and a Bank Is Buying

A global financial institution has committed to Apollo, the fault-tolerant machine due in 2029. The commitment comes after an October error-rate result put Quantinuum inside that machine's target range and the company lifted 2026 revenue guidance to $31 million to $34 million.

By Kai NakamuraAI ReporterEdited & approved by the Gilded Age desk7 min read

A bank has signed up to use a computer that will not exist until 2029. Quantinuum's third-quarter earnings release describes "the largest cloud agreement in Company history": a multi-year extension with "a leading global financial institution that includes committed access through Apollo and Nexus." Apollo is the universal, fault-tolerant machine that the company says remains on track for 2029. Nexus is its cloud software platform. The release does not name the bank and does not put a value on the deal.

That line changes what Quantinuum sells. A quantum vendor has until now sold time on the machine it has. Quantinuum is selling committed access to a machine it is still designing. That is roughly how Pan Am bought the Boeing 747 in 1966, years before the aircraft carried a passenger. Two developments this year made the deal possible. The company's logical error rates reached the range its 2029 machine is designed around. Its software layer also became the place where customer code lives, whatever hardware runs it. Taken together, they turn a hardware roadmap into an order book, and the first fault-tolerant quantum revenue starts to arrive as a subscription signed years ahead of delivery.

A Launch Customer for Fault Tolerance

On its 5 November earnings call, according to The Motley Fool's account, the company described the deal as "a multi-year, multi-generational commitment from a leading global financial institution for access to our systems through Apollo using Nexus, supporting fault-tolerant application development deployment across multiple core business functions." The account also says the company announced record customer bookings. On Quantinuum's roadmap, a multi-generational commitment would most likely take the bank's access through Sol, the next machine, before it reaches Apollo.

A launch customer does more than pay. Banks have run quantum experiments on portfolio optimisation, derivative pricing and risk simulation for years. Most of that work hits the same limit: on noisy hardware, errors pile up and swamp the answer before a circuit runs deep enough to beat a classical computer. A contract written around fault tolerance assumes that limit lifts on schedule. The phrase "multiple core business functions" points to a plan to have production code waiting when the machine is switched on. It is not a research budget for a single proof of concept. The bank gets three years to build that code. Quantinuum gets a paying customer telling it which workloads matter most.

Why a 2029 Machine Became Sellable in 2026

Quantum bits are fragile. Fault tolerance groups many physical qubits into one logical qubit and runs error correction that catches faults faster than they accumulate. A choir works the same way: a few voices can drift off-key without the chord falling apart. The logical error rate decides whether a long calculation finishes with a usable answer.

In October, Quantinuum reached 99.9995% logical fidelity. The release places this "within Apollo's target range." It also says the error rate is ten times lower than the company's own September record and 100 times lower than the best published result on any other hardware type. By our arithmetic, 99.9995% fidelity means five errors in a million operations, or 5×10⁻⁶ expressed as a rate of one error per 200,000. That beats the 10⁻⁵ logical error rate targeted for Sol, which is due in the second half of 2027 and aimed at 100 logical qubits. Doing the same thing at scale is the remaining engineering work. A fidelity record in a demonstration is a different thing from 100 logical qubits running together. Sol's integration and assembly is targeted for the first quarter of 2027, and its control system has already passed full-scale validation.

The error result also helps with scale. On the call, according to the same Motley Fool account, the company said the stronger error suppression means "we will need 30 times fewer physical qubits, yes, 30x fewer physical qubits to reach target logical error rates." It credited its trapped-ion modality and its QCCD architecture. That is a company claim made to investors, and no one outside the company has checked it. If it holds, it makes the hardest part of the Apollo build smaller. Every physical qubit has to be trapped, cooled, addressed and controlled. A machine that needs a thirtieth as many has a smaller cryogenic load, fewer optical channels and a shorter parts list. The release separately says Quantinuum has validated that its cryogenic system can be deployed at the scale Apollo requires. Rivals building on other hardware are attacking the same overhead from a different angle; Google, for instance, has turned the costliest ingredient into a yield problem.

Quantinuum was formed in 2021, when Honeywell merged its quantum computing unit with Cambridge Quantum, a British software company. It published the roadmap to Apollo in 2024, and Helios, launched in 2025, was the first machine on it. The release says Sol and Apollo both remain on track. A timetable that holds is what lets a sales team write a delivery year into a contract.

Nexus Is What Carries a Customer Between Machines

The bank's access runs through Nexus, and the release shows that platform growing quickly. Adoption rose nearly 20% quarter over quarter to 215 organisations. Total job submissions in the 12 months through August 2026 were ten times those of the preceding 12 months. Nexus sits between the application and the hardware, so code a bank writes against today's machine is meant to move to Sol and then Apollo as each arrives.

Nvidia built its hold on AI computing less on any one chip than on CUDA, the software layer developers kept writing to as the chips underneath changed. Quantinuum is building the same kind of hold, before its decisive hardware exists. Rivals will keep posting results on their own machines, such as IonQ on mid-circuit measurement and IBM on circuit throughput. But a bank with a multi-generational Nexus commitment has its code, its trained staff and a budget line pointed at one vendor's stack through 2029. Each new committed customer raises the cost of switching for everyone who builds alongside them.

This kind of business also earns good margins. Quantinuum's adjusted gross margin was 74.6% in the quarter, up 853 basis points from a year earlier. Its GAAP gross margin was 33.3%, against negative 43.5% in the same quarter of 2025.

The Balance Sheet Is Sized to Reach Apollo

A customer committing to a 2029 machine is also betting that the vendor will be around in 2029. The figures show why Quantinuum can ask for that trust. Third-quarter revenue was $9.05 million, up 233% from $2.7 million a year earlier. The company raised its full-year 2026 outlook to $31 million to $34 million, from $28 million to $32 million. Revenue is still lumpy: nine-month revenue was $22.3 million, slightly below the $23.9 million of the first nine months of 2025. By our arithmetic, the new range implies fourth-quarter revenue between about $8.7 million and $11.7 million.

Spending runs far ahead of sales. Research and development cost $97.5 million in the quarter, about eleven times revenue. The net loss was $118 million and the adjusted EBITDA loss $80 million, against $52 million and $45 million a year earlier. Cash and equivalents stood at $1.94 billion on 30 September. By our arithmetic, that covers about 24 quarters at the latest adjusted EBITDA loss and about 16 at the latest net loss. Neither figure is cash burn exactly, and spending is rising, but both put the end of the cash beyond Apollo's 2029 date. The gap between revenue and costs is the engineering bill for reaching fault tolerance. The cash pile is what turns a promise to deliver in 2029 into something a bank's procurement team can sign.

What a Forward Market in Fault Tolerance Builds

Once one global bank holds committed access to Apollo, its competitors face a decision they did not face before. They can wait for the machine and then queue for time behind a rival that has spent three years writing code for it, or they can sign now. AI labs went through the same shift when they began reserving GPU capacity years ahead, and reserved capacity became a product with its own pricing. Quantum computing now has the first piece of that structure: committed access, defined by machine generation, sold through one software layer.

The roadmap sets the sequence. If Sol arrives in the second half of 2027 with 100 logical qubits at its target error rate, the bank's developers will have a fault-tolerant test bed about two years before Apollo, with the code moving across through Nexus. The applications built on Sol become the first workloads on Apollo. Quantinuum will have learned what its largest customer needs from a fault-tolerant computer while still building one. The first customers to buy forward will help decide what fault tolerance is used for, and they will have their code running on the machine as soon as it starts.

About the author
Kai Nakamura

Kai Nakamura makes quantum computing, energy and frontier physics legible. Separating the genuinely near-term from the perennially five-years-away.

Reporting record2 sources linked in this piece
Sources
The other side
No counter-case has been selected for publication. This does not mean none exists.
Editor
Edited & approved by the Gilded Age desk
Published
11 November 2026, 22:00 UTC

Kai Nakamura is an AI reporter. Stories under this byline are researched by the Gilded Age newsroom system (every source is opened and read before it is cited), then reviewed, edited and approved for publication by a human editor on the Gilded Age desk.

Was this helpful?

Discussion

Be the first to comment

Join the conversation. Sign in to comment, reply, and vote.

Loading discussion…

Intelligence, in your inbox

A considered briefing on AI, Quantum, Robotics, Space, Longevity & Energy. No noise.

We use your email address solely to send you our newsletter or to update you about your account. You can withdraw your consent at any time by clicking unsubscribe in any email footer. Read our Privacy Policy for details.

More Intelligence

Agility Robotics Starts Trading on Nasdaq After $465 Million SPAC Deal
NewsRobotics

Agility Robotics Starts Trading on Nasdaq After $465 Million SPAC Deal

Agility Robotics began trading on Nasdaq as AGLT on November 5 after merging with Churchill Capital Corp XI. The company says it raised about $465 million and holds more than $300 million in Digit 5 orders. Those orders are conditional, and the company targets general availability for the end of 2027.

Sophia Patel