Nvidia and Jülich simulate a 50-qubit quantum computer on classical hardware

Nvidia and scientists at Germany’s Jülich Supercomputing Centre have completely simulated a 50-qubit universal quantum computer system on classical hardware. The partners explain it as the very first accomplishment of its kind.

The run occurred on JUPITER, Europe’s very first exascale supercomputer. The previous high-water mark was a 48-qubit simulation, embeded in 2019 by Jülich scientists utilizing Japan’s K computer system.

2 qubits, one large leap

The turning point was revealed in a news release dated November 11, 2025. Every qubit contributed to a simulation doubles the quantity of details the classical maker needs to track, so little gains in qubit count need outsized dives in memory and calculate.

For this run, that implied dealing with a state vector of more than 2 quadrillion complex amplitudes. JUPITER is constructed on Nvidia GH200 Grace Hopper Superchips, which combine a CPU and a GPU with hybrid memory gain access to, enabling processors to share a bigger swimming pool of memory than a GPU might reach by itself.

The nodes are looped with Nvidia’s Quantum-X800 InfiniBand networking.

The software application did a great deal of the heavy lifting

The Jülich Universal Quantum Computer Simulator, now updated to a variation called JUQCS-50, got a number of essential software application enhancements.

The very first was improved CPU unloading, which lets the system push parts of the work onto CPU memory when GPU memory runs brief. The second was an eightfold compression of memory requirements. The 3rd upgrade was smarter management of network traffic in between nodes. Taken together, the modifications provided a speedup of around 16.6 times compared to the previous 48-qubit simulations.

The job was supported by the JUPITER Research and Early Access Program (JUREAP) and is incorporated within JSC’s JUNIQ quantum user center.

Why mimic a quantum computer system at all

Physical quantum hardware is exceptionally conscious its environment, and mistakes sneak in continuously. A classical simulation acts like a best quantum computer system with no disturbance, making it beneficial for benchmarking quantum algorithms and screening mistake correction and fault-tolerance approaches.

What this suggests for the quantum race

For Nvidia, the outcome strengthens its positioning as main to quantum computing workflows without constructing a quantum computer system itself. The GH200’s hybrid memory style was a crucial active ingredient, using a concrete display for how memory architecture identifies how far a quantum simulation can scale.

For Europe, the run is an early evidence point for JUPITER, which was inaugurated in September 2025. A record-setting clinical outcome within approximately 2 months of launch shows the system’s ability as part of the exascale computing tier.

Since memory needs double with each included qubit, brute-force simulation can not keep up forever. The 2019 record took roughly 6 years to fall, and each brand-new record needs either drastically more hardware or cleverer compression.

Disclosure: This short article was modified by Diego Almada Lopez. For more details on how we produce and examine material, see our Editorial Policy.