Nvidia Expands AI Platform Support to RISC-V CPUs

Nvidia has announced its plans to expand support for RISC-V CPUs in its AI platform, adding a new option for server CPU choices alongside x86 and Arm.

The move was revealed by Nvidia senior engineer Frans Sistermans at the Hotchips 2026 conference held at Stanford University. RISC-V is an open-source CPU design method that allows anyone to use it.

Nvidia explained that it has prepared requirements for CUDA, a GPU development platform, and NVLink Fusion, an AI server connection technology, to operate on RISC-V-based CPUs.

The company also announced its cooperation with SiFive, a RISC-V company, as a new partner in the NVLink Fusion program. SiFive will hold a public demonstration of running CUDA on RISC-V-based CPUs for the first time.