Nvidia accelerates Google’s quantum AI design using quantum physics simulation

Nvidia is giving Google’s Quantum AI a boost.

Image credit: Nvidia.

Nvidia is working with Google Quantum AI in order to accelerate the design and development of its next-generation devices for quantum computing using simulations powered Nvidia.

Google Quantum AI uses the Nvidia Eos Supercomputer and the hybrid quantum-classical platform to simulate the physics for its quantum processors. The development of commercially viable quantum computers will only be possible if we scale
quantum hardware while keeping the noise in check, said Guifre Vial, a research scientist at
Google Quantum AI in a press release. “Using Nvidia-accelerated computing, we are exploring the noise
consequences of increasingly large quantum chip designs.” These simulations are prohibitively expensive to run. Google, however, can use the CUDA platform to employ 1,024 Nvidia Tensor Core GPUs on the Nvidia supercomputer Eos, to perform the largest and fastest dynamical quantum device simulations in the world — at a fraction the cost. Tim Costa, Nvidia’s director of quantum and HPC, said in a press release that “AI supercomputing will be helpful for quantum computing’s successful.” “Google’s use demonstrates the central importance of GPU-accelerated simulations in advancing quantum computation to solve real-world issues.”

With CUDA Q and H100 GPUs Google can perform fully detailed, realistic simulations for devices containing up to 40 qubits – the largest simulations performed of this type. CUDA Q’s simulation techniques allow noisy simulations to be run in minutes, instead of taking a week.

Software that powers these accelerated dynamic simulators will be made publicly available on the CUDA Q platform, allowing quantum hardware designers to rapidly scale up their system designs.