Oracle and Quantinuum have agreed a multi-year partnership under which Quantinuum will physically install its Helios quantum computer inside a US Oracle Cloud Infrastructure AI data centre, and Oracle will sell access to it as a managed OCI service.

No financial terms were disclosed, no launch customer was named, and no firm date was given beyond a preview “in the coming months”.

Most quantum cloud offerings are brokerage: a customer sends a job from a hyperscaler’s console to a machine sitting in the quantum vendor’s own facility, and the results come back.

Putting Helios inside an OCI data centre puts the quantum processor on the same fabric as Oracle’s GPUs and HPC capacity, under the same governance and security controls, which is what makes hybrid workloads, the ones that hand a problem back and forth between classical and quantum hardware, worth attempting at all.

Helios is not new hardware. Quantinuum launched it commercially in November 2025: a trapped-ion machine with 98 physical qubits, 48 demonstrated logical qubits, and an average two-qubit gate fidelity of 99.921%.

Its power draw is roughly 60kW, against the tens of megawatts a leading classical supercomputer consumes, though that comparison flatters the machine considerably given how narrow the set of problems it can currently address remains.

“Deploying Helios inside OCI gives Quantinuum and Oracle an opportunity to create a unique deeply integrated environment for hybrid workloads, explore enterprise use cases with customers, and accelerate commercial adoption,” said Rajeeb Hazra, Quantinuum’s president and chief executive.

Mahesh Thiagarajan, executive vice-president of Oracle Cloud Infrastructure, was more careful about what is on offer.

The goal, he said, is to give developers “a practical and secure way to explore how quantum computing could complement their existing AI and HPC workloads on Oracle Cloud Infrastructure while improving compute efficiency and energy use”.

The named target workloads are the usual list: drug discovery, materials science, financial modelling, and large-scale optimisation in logistics and energy.

Oracle is arriving late. AWS has offered Braket since August 2020, Microsoft’s Azure Quantum has been generally available since February 2022 and already carries Quantinuum’s own H-series hardware, and IBM has sold access to its own machines since 2016.

Oracle’s differentiator is not breadth but depth: one vendor, one machine, physically present, rather than a marketplace of remote QPUs. Whether enterprises want that trade is the thing the preview will establish.

Nvidia is absent from the announcement, despite Quantinuum having an existing NVQLink integration dating to last November, and despite Oracle’s hybrid pitch resting on GPU proximity.

For Quantinuum, the deal is the second significant validation of a year that has already reshaped the company.

It priced its Nasdaq IPO in June at $60 a share, selling 28 million Class A shares to raise $1.68bn at a valuation of roughly $14bn, above its own upsized range and more than twenty times oversubscribed, on revenue of $30.9m and a net loss of $192.6m for 2025.

It trades as QNT. Its roadmap runs to Sol in 2027 and Apollo, the machine it says will demonstrate genuine quantum advantage, in 2029.

Europe has been building in a different direction. Finland’s IQM has put superconducting machines directly into research data centres rather than routing access through a hyperscaler, on the argument that co-location with classical HPC is where useful work will first appear.

Oracle’s arrangement borrows that logic and applies it to a commercial cloud, which is either the sensible synthesis or a way of catching up quickly, depending on how generous you are feeling.

Investors were unmoved as Oracle closed down 3.69% at $145.48 on Tuesday and Quantinuum slipped 0.94% to $56.06, recovering slightly after hours.

A partnership with no disclosed terms, no named customer, and no launch date is difficult to price, which is roughly the position quantum computing has occupied for a decade.

What none of the parties has offered is a workload that runs better this way today. The release leans on the phrase “explore enterprise use cases with customers”, which is the industry’s standard construction for work that has not yet produced a result.

Quantinuum’s own $30.9m of annual revenue against a $192.6m loss is a reasonable proxy for how much commercial quantum computing currently exists anywhere.

Get the TNW newsletter

Get the most important tech news in your inbox each week.