On October 19, 2026, NVIDIA announced the successful execution of a fault-tolerant quantum computing simulation using the ‘JUPITER’ system, Europe’s first exascale supercomputer located at the Jülich Supercomputing Centre in Germany. This research, conducted at a 104-qubit scale, represents the largest simulation to date that comprehensively models the complex behavior and noise interference of actual quantum processors.
104-Qubit Simulation Executed on Europe’s First Exascale Supercomputer
According to NVIDIA’s press release, the computation leveraged 8,192 NVIDIA GH200 Grace Hopper Superchips installed in the JUPITER supercomputer. The research team modeled in detail how a 104-qubit-based processor performs fault-tolerant operations under realistic hardware error and noise conditions.
While previous research at the Jülich Supercomputing Centre simulated 50 qubits in an ideal, noise-free environment, this study incorporates hardware-level noise. This provides critical data to estimate design performance and error patterns with high precision before prototyping actual quantum hardware.
90% Reduction in Qubits and Precision Validation of Noise Impacts
A key achievement of this simulation is the validation of ‘Magic State Cultivation,’ a critical process for fault-tolerant quantum operations. Fault-tolerant computing, which includes quantum error correction, continuously consumes special qubit states known as magic states during computation.
NVIDIA researchers simulated the latest magic state cultivation techniques under noisy conditions, demonstrating that fault-tolerant operations remain valid while reducing the required number of physical qubits by 90% compared to traditional methods. This high-precision simulation allows for the reduction of trial-and-error in hardware and enables architectural improvements prior to physical prototyping.
Research Workflow Based on NVAQC and CUDA-Q Platform
Before scaling the computation to 8,192 GPUs on the JUPITER supercomputer, the team conducted development and pre-testing on a GB200 NVL72 system located at the NVIDIA Quantum Cloud (NVAQC). By optimizing algorithms on this next-generation infrastructure, the research validation period, which could have taken weeks, was shortened to just a few days.
NVIDIA cuQuantum software was used to accelerate the simulation, while the NVIDIA CUDA-Q platform was utilized for fault-tolerant logical circuits and circuit representation. NVIDIA stated that this technology will contribute to accelerating the commercialization of advanced quantum computing in fields such as drug discovery, financial modeling, and materials science.
This article is based solely on one official press release from NVIDIA (October 19, 2026) and does not include independent verification from other institutions.
Official Source
- NVIDIA Newsroom: NVIDIA Runs Record Simulation of Fault-Tolerant Quantum Computing at Jülich Supercomputing Centre (Published October 19, 2026)