Nvidia adds RISC-V instruction set architecture support to CUDA programming ecosystem.

Nvidia announced official support for CUDA on RISC-V CPUs during the RISC-V Summit in China on Friday. CUDA is Nvidia's high-level software abstraction layer that enables applications to interact with its GPUs; without CUDA support on a CPU architecture, GPU functionality is limited at best. The announcement positions RISC-V processors based on the open instruction set architecture to serve as host CPUs for Nvidia GPUs, expanding the company's stated goal that "our goal is to support all major architectures with CUDA. This includes x86, Arm, and now RISC-V," according to a company spokesperson.

The timing of the announcement in China reflects the nation's strategic push to reduce reliance on Western processors, with RISC-V playing a central role in that effort. Alibaba's R&D division XuanTie unveiled the C930 CPU core in March, targeting server, PC, and automotive applications. Meanwhile, the Xiangshan project has developed a high-performance RISC-V processor core that the team claims comes within spitting distance of Arm's two-year-old Neoverse N2 cores. This development coincides with Nvidia's resumption of shipments of its China-spec H20 accelerators to China.

Nvidia's expansion to RISC-V builds on existing experience across multiple processor architectures. CUDA already runs on both x86 and Arm64-based processors, including Nvidia's in-house Grace CPUs. Prior to Arm support, CUDA was supported on IBM Power-based systems; the Department of Energy's Sierra and Summit supercomputers paired Power9 processors with V100 GPUs. Nvidia has also embedded RISC-V cores in its GPUs for years—RISC-V International estimates that Nvidia shipped more than a billion RISC-V cores in 2024 alone, with anywhere from 10 to 40 cores in every GPU sold, handling low-level functionality like video codecs, chip-to-chip interconnects, power management, and security.

Implementation approaches remain under consideration. Nvidia could develop a RISC-V single board computer (SBC) combining RISC-V cores with an Nvidia GPU, mirroring the Arm-based SBCs it currently offers. Alternatively, Nvidia could partner with RISC-V-based chip or server vendors to build reference designs similar to its 2019 partnership with Ampere. The company indicated that timing for RISC-V adoption will depend on the availability of HPC-ready systems using the instruction set architecture.