NVIDIA runs 104-qubit fault-tolerant quantum simulation at Jülich

By Splitfeed EditorialLast updated Oct 19, 2026, 19:00 UTC

On October 19, 2026, NVIDIA announced it had run a 104-qubit simulation of a fault-tolerant quantum computing protocol on the JUPITER supercomputer, hosted at the Jülich Supercomputing Centre in Germany. JUPITER holds the distinction of being Europe's first exascale-class machine, meaning it can perform at least one quintillion floating-point operations per second. The quantum approach underlying the simulation was designed to require 90% fewer physical building blocks than earlier fault-tolerant methods, a claim NVIDIA made through its official newsroom on the same date.

Fault-tolerant quantum computing is widely considered the threshold that separates experimental curiosity from practical utility. Real quantum hardware is fragile, prone to errors that accumulate quickly, and the overhead required to correct those errors has historically demanded enormous numbers of physical qubits. A method that cuts that overhead by 90% would, if it holds up under scrutiny, significantly lower the resource requirements for reaching useful quantum computation. Running such a protocol at 104 simulated qubits on classical supercomputing hardware is a way of stress-testing the approach before committing to physical quantum hardware, and doing it at this scale on an exascale machine represents a new benchmark. What remains genuinely in dispute is how closely classical simulation of fault-tolerant protocols predicts real-world quantum performance, and whether the efficiency gains claimed for this particular approach survive contact with actual quantum hardware noise and physical constraints.

No reactions from the Pro-AI, Anti-AI, or Middle Ground camps had been published at the time of this story. Typically, the Pro-AI camp would frame a result like this as concrete progress toward transformative quantum-classical hybrid computing, pointing to the efficiency improvement as evidence that the field is maturing faster than critics expect. The Anti-AI and skeptical tech camp would likely question whether classical simulation tells us much about real quantum advantage, and would flag the distance between a supercomputer modeling a protocol and actually running it on quantum hardware. A middle-ground perspective would probably credit the methodological advance while cautioning that the gap between simulation and physical demonstration remains substantial.

The argument will sharpen once researchers attempt to reproduce the 90% efficiency claim on physical quantum hardware, or when peer-reviewed publication of the underlying protocol allows independent verification of the building-block reduction figures. Any experimental demonstration on actual qubits, even at smaller scale, would be a meaningful next data point.