Nvidia Unveils Open-Source GPU-Native Storage Tech

Nvidia has open-sourced its cuFile storage stack, making GPU-native storage access an open industry standard. The company has released the code under the new Accelerated IO Special Interest Group (SIG) on GitHub, with Google, Intel, and Meta as founding maintainers.

The cuFile APIs and complete software stack have been made available to the public, marking a major shift from proprietary technology to community-driven open-source project. This move enables storage vendors and AI infrastructure providers to adopt GPU-native storage across the industry.

At the center of the release is SCADA (Scaled Accelerated Data Access), a framework that moves storage control directly onto the GPU, reducing storage latency to microseconds. Nvidia claims this technology strengthens Linux interoperability and security by using standard Linux privilege mechanisms for direct storage access.

Nvidia demonstrated the technology on a single server equipped with 44 Gen6 SSDs, achieving 230 million random 512-byte reads per second while the CPU remained idle. The company also launched Storage-Next, an ecosystem initiative involving more than 40 companies to advance open GPU-native storage technologies.