Nvidia launches Quantum Cloud to simulate quantum processors on GPUs

True quantum computers are still too noisy and rare for real-world use, so Nvidia came up with a practical shortcut: just fake them with GPUs.

At its annual GTC conference, Nvidia unveiled Nvidia Quantum Cloud, a new cloud microservice currently in early access that allows researchers to simulate quantum processing units (QPUs) on classical GPU clusters.

Simulating a future that hasn't arrived

Building physical quantum hardware is notoriously tricky. Existing QPUs lack the scale and stability needed to test production software, meaning developers often hit a wall before writing a single line of viable code.

Nvidia is stepping into the gap with its cuQuantum library and CUDA-Q hybrid programming platform, which it claims is already used by roughly three-quarters of organizations deploying physical QPUs. Quantum Cloud will be rolled out across major cloud giants, including AWS, Google Cloud, Microsoft Azure, and Oracle Cloud, alongside dedicated quantum players like IonQ, IQM, and Quantinuum.

Early adopters are already putting the infrastructure to work. HSBC used CUDA-Q and cuTensorNet software to simulate 165 qubits on GPUs to build a quantum machine learning algorithm for fraud detection, breaking previous scale records. Meanwhile, researchers at the University of Toronto are fusing large language models with quantum eigensolvers to study molecular energies.

To power these workloads at scale, supercomputers are loading up on silicon. Japan's National Institute of Advanced Industrial Science and Technology is building the ABCI-Q supercomputer, which integrates more than 2,000 Nvidia H100 GPUs across 500 nodes to run high-fidelity simulations.

Selling shields before the spear lands

Nvidia is also tackling the security nightmare that comes with quantum computing: the inevitable cracking of modern encryption algorithms.

The company introduced the cuPQC library, designed to accelerate post-quantum cryptography on GPUs. In benchmarks run by the Open Quantum Safe project, executing the Kyber encryption algorithm on an Nvidia H100 GPU delivered a 500-fold speedup compared to a traditional CPU. Nvidia noted that its upcoming Blackwell architecture will push that performance even further, just as NIST prepares to release the first post-quantum security standards later this year.

Why it matters

For enterprise teams and researchers, Nvidia Quantum Cloud removes the painful friction of waiting for reliable physical quantum hardware. Instead of gambling on fragile experimental QPUs, developers can write, debug, and validate hybrid algorithms right now using familiar cloud tools and GPU capacity.

Strategically, this is classic Nvidia brilliance. Rather than fearing that quantum computing might someday make GPUs obsolete, the company is positioning CUDA-Q as the mandatory bridge connecting classical supercomputing to quantum machines. By owning the developer tools, simulation layer, and post-quantum defense software, Nvidia ensures that wherever quantum research happens, a rack of its GPUs will be billing hours in the background.

The catch is that classical simulation hits a hard mathematical wall. Simulating qubits demands exponential amounts of classical memory, meaning GPUs cannot mimic large-scale quantum machines forever. But until fault-tolerant QPUs become commercially viable, Nvidia has successfully turned quantum computing into yet another reason to buy more graphics chips.