Third quarter 2026 earnings call, November 5, 2026

Quantinuum used its third-quarter call to argue that its lead in fault-tolerant hardware is widening, and the numbers it offered are specific. Management said a logical-qubit experiment on its Helios system in October reached 99.9995% fidelity, equal to 5 errors in 1 million operations. CEO Rajeeb Hazra said this was 10x lower than the company's own September record and 100x lower than Google's, which had published the previous best result across hardware modalities. These are company-reported figures, and the comparison is against Google's superconducting work. Even so, the claim is central to the investment case, because Quantinuum's roadmap to a universal, fully fault-tolerant machine called Apollo in 2029 rests on error correction that scales efficiently.

Error correction efficiency matters more than the headline fidelity

The efficiency of the error suppression is arguably the more consequential disclosure. Hazra said that each 2-step increase in code distance cut logical memory errors by a factor of roughly 12, against roughly 2 in Google's published superconducting result. Management translated this into a requirement for 30x fewer physical qubits to reach target logical error rates. Hazra credited the result to the company's trapped-ion modality and its two-dimensional QCCD architecture, in particular the all-to-all connectivity. In his words, the company achieved "an exponential reduction in error correction for a reasonably small amount of increase in the number of physical qubits that are used to create a logical qubit."

In response to analyst questions, Hazra said the results were produced with a universal gate set that includes non-Clifford gates, which is where other modalities have struggled because of magic state overhead. He said Quantinuum's 2025 magic state work was slightly better than Google's reported 40x reduction in magic state errors, with a significantly higher acceptance rate. A higher acceptance rate means fewer discarded states, less physical qubit consumption and less time spent building them. Management did not publish logical specifications for Lumos, the successor to Apollo, but said it expects to keep applying these techniques "from the scale of Helios to a million-plus qubit machine."

Roadmap: Sol on track for 2H27, with more confidence than Helios had

Hazra said Sol, the 100-logical-qubit system targeting a logical error rate of 10^-5, remains on track for commercial availability in the second half of 2027. He said management has greater confidence in this timeline than it had for Helios at the same stage. In the quarter the company validated Sol's full control system for moving qubits in the two-dimensional trap at product scale. Helios managed a single qubit-pair crossing, whereas Sol must coordinate close to 100 pairs. Full system integration and assembly should finish in the first quarter of 2027, followed by validation, including running real jobs.

On Apollo, Hazra identified three subsystems being de-risked: the qubit transport system, the laser and optics system being prototyped with Monarch Quantum, and the cryogenic system. He noted that Quantinuum does not need cryogenics to make qubits, only to maintain vacuum. The closed-loop recovery of helium nevertheless affects operating costs, and "a percentage or 2" of gain in round-trip recovery changes the machine's total cost of ownership. The company finalized its CHIPS R&D award in September and was the only trapped-ion company selected. The award is $100 million, with the U.S. government receiving roughly 2 million shares.

Financials: strong growth from a small base, with heavy losses

Third-quarter revenue was $9.1 million, up 233% year over year, driven by the cloud platform along with research and services. Government accounted for the majority of revenue. Non-GAAP gross margin was 74.6%, which CFO Nitesh Sharan said reflects a heavy cloud mix and will move with product mix as hardware sales grow. Bookings in the quarter were $98 million, including the Oracle deal, bringing year-to-date bookings to roughly $104 million. Management expects more than $120 million for the full year.

The cost base is scaling much faster than revenue. R&D was $97.5 million, including $27 million of stock-based compensation, versus $42.4 million a year ago. Sales and marketing was $14.9 million versus $4 million. G&A was $22.1 million versus $5.3 million, reflecting public-company costs. Adjusted EBITDA was a loss of $80.5 million, compared with a loss of $45.3 million a year earlier. Total GAAP net loss was $117.9 million, of which $17 million, or $0.46 per share, was attributable to Quantinuum Inc. after non-controlling interest. Non-GAAP EPS was a loss of $0.29. Cash was more than $1.9 billion with no debt, which management called more than sufficient to fund the roadmap. At the current quarterly burn this provides a long runway, though the cost lines are rising quickly. Sharan said the CHIPS award is classified as an equity contract, so it does not affect the P&L.

Guidance raised for 2026, with 2027 growth above 100% reiterated

The company raised its 2026 revenue outlook to $31 million-$34 million from $28 million-$32 million. It reiterated expected 2027 growth of more than 100%, now off the higher base. Hazra described the raise as "broad-based" across governments, national labs, universities and enterprises, and across the U.S., Europe and Asia. Asked how much of 2027 is covered, he said "a good amount," without quantifying it. He said the existing backlog gives "a fair amount of confidence" in at least 100% growth, that Oracle is important but not the only driver, and that formal 2027 guidance will come on the next call. The absolute revenue base remains small, so the 2027 target implies revenue of roughly $65 million-$70 million at the guidance midpoint, set against a quarterly adjusted EBITDA loss above $80 million.

Commercial traction: largest cloud deal and a second hyperscaler

Quantinuum signed what it called its largest-ever cloud agreement, a multi-year, multi-generational commitment through Apollo from a leading global financial institution, using Nexus for fault-tolerant application development. Hazra said customers are signing before fault tolerance arrives for two reasons: early engagement that maps use cases to each hardware generation, and the predictability of the roadmap. He said the work has shifted from proof of concept to "pre-building of the application" so that no gap opens between hardware readiness and business value. Sharan added that the financial institution is expanding what it spends with Quantinuum, and that this pattern is showing up across a number of customers.

The company also signed an agreement with an unnamed hyperscaler, adding to Oracle. The economics of the new arrangement are not settled. Sharan said "we're still working through some pieces" of the revenue model and described the deal as "a go-to-market access play." Hazra distinguished the two structures. Oracle bought the system and the full stack, while in other arrangements Quantinuum is paid for value delivered to customers already on the hyperscaler's platform. He said the company monetizes the full stack, including Nexus and its chemistry libraries, rather than "commoditizing the hardware." The Helios deployment into Oracle's AI data center remains on track.

Nexus gains users and a fault-tolerance roadmap

Organizations building on Nexus rose to 215 from 180 last quarter, and job submissions rose 10-fold in the 12 months through August versus the prior 12 months. The company announced a multi-generational roadmap that aligns the software with each hardware generation through Apollo. The September release extended the Guppy programming language to logical qubits and supported error correction codes. A December release will add hybrid classical-quantum workflows and execution analytics. Hazra said the Apollo-era Nexus will let developers choose error correction codes "like a library" and may recommend codes for a given application. He framed Nexus as sticky because developers bind their applications to the language and libraries. About 215 organizations is still a small base, and most run on Quantinuum's own cloud.

Government programs and international systems

Quantinuum is in active discussions with DARPA on a Stage C scope of work while finishing Stage B of the Quantum Benchmarking Initiative. Morgan Stanley's Joe Moore noted that some others have already passed through Stage C. Hazra gave no timeline and called DARPA "an important validation confidence builder" while stressing that the company's roadmap is also shaped by enterprise customers. The absence of a Stage C award is the notable gap in the government narrative. Separately, the company is part of four winning university proposals in Phase 1 of the Department of Energy's Genesis mission. A second Helios system is on site in Singapore, with commissioning expected next spring, and an upgraded system at RIKEN in Japan is being commissioned next week.

Strategy and capital allocation

Asked about M&A and multimodality, Hazra declined to speculate beyond saying the company is "always actively looking at M&A" with its balance sheet, with software the stated focus. He expressed full confidence in trapped ions and argued that time to solution depends on accuracy as well as speed, "because there's no point in getting the wrong answer quickly." Management also pointed to an IP-based revenue stream: it owns a workflow for metal-organic frameworks that it is already monetizing in a couple of pharma engagements. This is a possible route to outcome-based revenue in work such as the Johnson & Johnson collaboration, though it remains early. The company is also promoting QOPs, a benchmark developed with Sandia National Laboratories and NVIDIA, which it hopes becomes an industry standard.