Exascale Muscle Meets Quantum: Inside the Record 104-Qubit Simulation
Europe’s first exascale supercomputer, JUPITER, installed at the Jülich Supercomputing Centre in Germany, has been used to carry out the largest comprehensive simulation of a fault-tolerant quantum computer yet attempted, modelling 104 qubits end to end. The run is a landmark not because a classical machine “beat” a quantum one, but because it gives researchers a trustworthy reference against which real quantum hardware can be checked.
Why simulate a quantum computer at all?
Quantum processors are notoriously difficult to verify. Their outputs are probabilistic, their errors subtle, and their inner states impossible to inspect directly. A full classical simulation lets scientists compare a physical device’s behaviour against an ideal mathematical model, exposing where noise creeps in and which error-correction schemes actually hold up. As our explainer on newsross.com notes, verification is becoming the bottleneck of the quantum era.
Simulating qubits classically is brutally expensive: every additional qubit doubles the memory required. Reaching 104 qubits in a comprehensive fault-tolerance simulation therefore demands the memory bandwidth and interconnect speed that only an exascale system, performing over a quintillion operations per second, can supply.
What JUPITER brings to the table
JUPITER pairs tens of thousands of NVIDIA GH200 Grace Hopper superchips with a booster module designed for AI and simulation workloads. Its architecture is built for exactly this kind of tightly coupled mathematics, where processors must constantly exchange data rather than work in isolation. The machine also anchors Europe’s wider push for technological sovereignty in high-performance computing.
Researchers say the simulation framework will be used to test quantum error-correction codes under realistic noise, helping hardware teams decide which designs deserve scarce fabrication resources. Agencies such as the U.S. Department of Energy pursue similar programmes on their own exascale systems.
A reference point for the quantum race
The deeper significance is strategic. As quantum hardware claims grow bolder, independent classical benchmarks keep the field honest. A simulation of this scale effectively raises the bar any quantum device must clear to demonstrate genuine advantage.
For now, the record stands as both a scientific tool and a statement: classical supercomputing is not being replaced by quantum machines just yet. Instead, the two are evolving together, each making the other more useful.
The energy question behind the power
Exascale machines draw as much electricity as a small town, and JUPITER was designed with that reality in view. Its direct liquid cooling and dynamic power management aim to keep performance per watt among the best in its class, a metric that increasingly decides where the next generation of machines gets built. A simulation that once would have taken weeks of queue time can now be scheduled, run and repeated, which changes how research groups plan their experiments: iteration, not access, becomes the limiting factor.
How error correction gets stress-tested
Fault-tolerant designs rely on encoding each logical qubit across many physical ones, then continuously measuring for errors without disturbing the computation. Simulating that cycle at 104 qubits means tracking an enormous web of entangled states and syndrome measurements, which is why the run is described as comprehensive rather than partial. Weak points show up as error rates that refuse to fall as codes scale, and those are exactly the findings hardware teams need early.
The results also feed a practical market: cloud providers selling access to early quantum hardware use classical benchmarks to set honest expectations for customers, and insurers and investors use them to price risk in a field prone to exuberant claims.
What this unlocks next
The immediate beneficiaries are algorithm researchers, who can now test error-correction ideas at scales that were previously reachable only in simplified form. Each verified result sharpens the targets for hardware teams and helps funding bodies distinguish credible roadmaps from optimistic ones. Over time, the same simulation capability improves the compilers and calibration tools that any future quantum cloud service will depend on, which is why national laboratories treat it as infrastructure rather than a one-off experiment. The virtuous circle, classical machines improving quantum ones that will one day accelerate classical problems, is becoming the defining dynamic of advanced computing.
Related reading: Who Watches the Agents? Microsoft Builds a Guard for Autonomous AI
