According to Tom's Hardware the move lets RISC-V step into a job that’s long been reserved for x86 or Arm, at least on the CPU side of CUDA-driven systems.

The announcement came during a keynote at the event delivered by Nvidia hardware engineering vice president Frans Sijsterman. He confirmed that CUDA components are being adapted to work seamlessly on RISC-V, meaning these open instruction set processors can finally orchestrate Nvidia GPUs without relying on Arm or x86 hosts.

Nobody’s betting on RISC-V storming hyperscale datacentres tomorrow, but it does make sense for CUDA-enabled edge devices like Nvidia’s Jetson modules.

The company’s slides showed GPUs crunching parallel workloads while a RISC-V CPU manages CUDA system drivers, application logic, and the operating system. A DPU handled networking, completing a setup that clearly screams heterogeneous compute with RISC-V at the centre of the orchestration.

The workloads are almost certainly AI-heavy, although Nvidia didn’t confirm it. The bigger takeaway is that Nvidia has just tied its proprietary CUDA stack to an open CPU architecture that’s developing at breakneck speed in China.

It’s not hard to see why. With restrictions keeping flagship GB200 and GB300 hardware out of China, Nvidia needs other ways to keep CUDA sticky. By supporting RISC-V, it opens the door for Chinese companies building custom silicon to stay within the CUDA ecosystem.

The move also strengthens options for Jetson developers who target embedded or specialised computing platforms. But there’s a deeper implication. If Nvidia is willing to let RISC-V into the CUDA ecosystem, it hints at a possible diversification beyond its usual closed host platforms.

Whether this is the start of a bigger shift in AI and HPC processor design or just a way to keep its hooks in Chinese markets remains murky. What’s clear is that RISC-V just gained a serious endorsement from the biggest name in GPUs, and that could push others to follow.