NVIDIA (NVDA) stock rose 1.54% to $217.43 on Monday after the company reported a major quantum computing simulation milestone in Europe. Shares gained $3.30 during the session, although intraday selling trimmed earlier advances. The announcement centered on a 104-qubit test aimed at improving fault-tolerant quantum computing.

Record 104-Qubit Simulation on JUPITER

NVIDIA researchers ran the simulation on JUPITER, Europe's first exascale supercomputer—a class of machine capable of at least a quintillion calculations per second—at Germany's Jülich Supercomputing Centre. The project used NVIDIA GH200 processors to model complex quantum operations under realistic hardware noise. According to NVIDIA, the work represents the largest comprehensive simulation of a fault-tolerant quantum processor.

Simulations of this kind fill a practical gap: fault-tolerant quantum machines at this scale have not yet been built, so classical supercomputers give researchers a way to test processor designs before committing to physical hardware.

The researchers examined magic-state cultivation, a method that helps quantum systems carry out reliable calculations. Quantum computers are vulnerable to errors because their qubits respond to noise during processing, and fault tolerance describes a machine's ability to keep producing accurate results despite those errors. Scientists therefore need techniques that limit these errors without requiring excessive hardware.

An earlier research effort at Jülich modeled 50 qubits under ideal conditions without noise. The new project extended that work by examining 104 qubits in a more realistic setting, helping researchers assess how proposed quantum machines might perform outside controlled simulations.

Cultivation Method Targets Lower Hardware Requirements

NVIDIA said the cultivation approach could require 90% fewer qubits than previous methods. The reduction matters because useful quantum computers need substantial resources to correct processing errors. Lower hardware requirements could simplify future designs and reduce the cost scaling them.

Magic states support calculations that ordinary error-correction methods cannot complete alone. However, quantum processors continually consume these states during fault-tolerant operations, which is why researchers study faster and more efficient ways to produce them.

The team tested how noise affects cultivation and identified ways to evaluate future improvements. Its findings could guide hardware developers before they commit resources to physical prototypes—the stage at which the projected qubit savings would first face conditions beyond simulation. The testing methods also support research in drug discovery, financial modeling, and advanced materials.

NVIDIA Expands GPU Simulation Capacity

The team first developed its methods using NVIDIA's GB200 NVL72 system at its Accelerated Quantum Research Center. Researchers expanded testing from four GPUs to hundreds before moving to JUPITER, where they eventually used 8,192 GPUs to perform the larger simulation.

That progression helped the team verify accuracy and compare overall performance across computing environments. It also shortened work that previously required weeks into a process lasting days, while preparing the software for JUPITER's much larger computing resources.

NVIDIA used its cuQuantum software to improve simulation performance across the entire system. Meanwhile, its CUDA-Q platform provided tools for describing fault-tolerant quantum circuits. Together, the two platforms demonstrated how conventional GPU supercomputers can support quantum research before engineers build larger physical processors—a role that lets NVIDIA's existing GPU and software products serve quantum development while such hardware remains in the research phase.

This article was originally published on Blockonomi.