Quantum computers usually live behind glass in a handful of specialist labs owned by the people who built them. This week Oracle decided to move one into the same building as its GPUs. On Monday the cloud company and Quantinuum announced a multi-year partnership to install Helios, a 98-qubit trapped-ion machine, inside a US-based Oracle Cloud Infrastructure AI data centre, where customers will reach it alongside Oracle’s high-performance computing and GPU kit.
Quantinuum is calling this its first on-premises deployment inside a major hyperscale cloud provider’s AI data centre, and the phrase to underline is on-premises. Instead of piping jobs off to a remote quantum lab, Oracle wants the quantum machine, the classical supercomputing and the AI accelerators sharing one floor, so a single hybrid workload can move between all three without leaving the building.
What Helios actually is
Helios is the sharp end of Quantinuum’s hardware. The 98-qubit system delivers 48 logical qubits and a two-qubit gate fidelity of 99.921%, the sort of accuracy figure that decides whether a quantum calculation is useful or noise. Quantinuum bills it as the most accurate commercial quantum computer available, and the fidelity number backs the boast up rather than leaning on qubit count alone.
It is also surprisingly frugal. Helios draws around 60 kilowatts, not counting external cooling, which Quantinuum says is under one percent of the power a top-tier supercomputer burns through. In an era where AI data centres are measured by their appetite for electricity, a quantum machine that sips rather than gulps is a decent selling point.
Why Oracle wants it next to the GPUs
The pitch is hybrid computing. Some problems, molecular simulation for drug discovery, certain financial-modelling and optimisation tasks, have pieces a quantum processor may handle better than any classical chip, while the rest of the job still runs faster on GPUs and CPUs. Keeping all of it in one data centre cuts the latency and the plumbing headaches of shuttling data between a cloud and a distant quantum lab.
“We believe the next phase of enterprise computing will be shaped by bringing quantum, AI, and high-performance computing together,” said Dr Rajeeb Hazra, Quantinuum’s president and chief executive. Deploying Helios inside Oracle’s infrastructure, he added, gives both firms a chance to “explore enterprise use cases with customers, and accelerate commercial adoption.”
Hype check
Worth keeping the feet on the ground. Nobody flipped a switch this week. The companies gave no firm launch date beyond a preview of the OCI quantum service “in the coming months,” and hybrid quantum-AI remains an experiment more than a product line. Quantinuum, which trades on the Nasdaq under the ticker QNT, is selling a vision of what enterprise computing looks like in a few years, not a finished tool a bank can plug in tomorrow.
Still, the direction of travel is real. Putting a quantum machine physically inside a hyperscaler’s AI data centre, rather than treating it as an exotic service reached over the internet, is a meaningful shift in how the industry expects this technology to be used. Drug researchers and quants will not care where the qubits live, only whether the answers come faster. This deal is a bet that the shortest path to that runs through the same room as the GPUs.
Did you know: trapped-ion machines like Helios encode qubits in individual charged atoms held in place by electromagnetic fields, a very different approach from the superconducting chips Google and IBM are known for.
Sources
Related on Top Tool Stack: Google’s Gemini 3.7 Flash Lands Cheap and Fast as the Flagship Slips · Cool Inflation Prints Push the S&P to a Record and Corner the Fed