NVIDIA says JUPITER ran record fault-tolerant quantum simulation

News

NVIDIA said JUPITER ran a 104-qubit fault-tolerant quantum simulation, modeling magic-state cultivation under realistic noise.

NVIDIA said on Oct. 19, 2026, that researchers used JUPITER at the Jülich Supercomputing Centre to run what it described as the largest comprehensive simulation of a fault-tolerant quantum processor to date. The company said the work modeled a 104-qubit system and was aimed at validating magic-state cultivation, a fault-tolerant approach that NVIDIA says could use 90% fewer physical qubits than previous methods.

Evidence and context

According to NVIDIA’s press release distributed through GlobeNewswire on Oct. 19, 2026, the simulation ran on JUPITER, described as Europe’s first exascale supercomputer and the largest in Europe, using 8,192 NVIDIA GH200 GPUs. NVIDIA said the project modeled how a realistic quantum processor performs fault-tolerant computation under hardware noise, rather than in idealized noiseless conditions.

The company said the result provides researchers with data to test future quantum designs before hardware is built, potentially reducing reliance on trial-and-error development. NVIDIA also said the simulation helps quantify the impact of noise on fault-tolerant systems and supports applications that could eventually matter in drug discovery, financial modeling and materials science. These are company-reported claims, not independently verified performance conclusions.

NVIDIA said the simulation work was developed and tested on a GB200 NVL72 system at its Accelerated Quantum Research Center before scaling to JUPITER. The company said that process helped reduce a weekslong simulation effort to days, but it did not provide a standardized benchmark or a precise end-to-end runtime.

Kristel Michielsen, director of the Jülich Supercomputing Centre at Forschungszentrum Jülich, said in NVIDIA’s release that JUPITER gives researchers the ability to investigate questions previously beyond reach and that the collaboration is intended to help scientists make better-informed decisions about future quantum designs. Jensen Huang, NVIDIA’s founder and chief executive, said the work shows how quantum processors and GPU supercomputers can be used together to tackle research that would otherwise be out of reach.

The announcement follows NVIDIA’s Sept. 14, 2026 update on its open-source CUDA-Q platform for fault-tolerant quantum computing and its Oct. 8, 2026 commitment to invest $1 billion to advance U.S. science over the next five years. The company’s latest release centers on simulation, not the operation of a physical fault-tolerant quantum computer.