Nvidia announced at the 2025 RISC-V Summit that its CUDA software platform is now compatible with RISC-V-based processors. A presentation by Frans Sijsterman, Vice President of Hardware Engineering at Nvidia, at the event held in China, detailed how the RISC-V architecture integrates with the CUDA ecosystem.

Nvidia Announces RISC-V Support for the CUDA Platform

Until now, the CUDA platform has worked on x86 and Arm-based processors. With this new announcement, the platform is now compatible with the RISC-V architecture, which stands out for its open-source and license-free architecture.

In the architectural model shown at the presentation, GPUs handle parallel processing tasks, while RISC-V-based processors manage the application logic, CUDA drivers, and operating system. The system also includes DPUs, which handle tasks such as data movement and network management. This architecture was described as part of Nvidia’s vision for heterogeneous computing.

RISC-V is not expected to replace common architectures like x86 or Arm in the short term. However, Nvidia states that RISC-V and CUDA can be used together in modules used in embedded systems like the Jetson.

This opens up new possibilities for customized hardware solutions and resource-limited edge devices. The ability for CUDA to work with RISC-V opens up alternative development paths, particularly for developers working on artificial intelligence and embedded systems.

The announcement has significant regional implications in addition to its global implications. This announcement, made at an event in China, highlights the rapidly growing interest in RISC-V in the country in recent years and Nvidia’s efforts to reshape its position in this market.

Nvidia, which has been unable to sell high-end AI hardware like the GB200 and GB300 to China due to US government export restrictions, aims to fill the gap created by these restrictions with RISC-V-enabled solutions. This will allow the company to maintain its presence in the Chinese market while preserving the CUDA ecosystem.

This step by Nvidia is seen as a sign of a transition that will further enhance the role of systems running on open hardware architectures in the field of artificial intelligence. Following this development, the technical details and development tools for RISC-V-supported CUDA applications are expected to be expanded in the coming period.