Nvidia buys AI developer to expand open-source AI push

Nvidia has acquired AI software firm SchedMD as the chip designer doubles down on open-source technology and steps up investments in the artificial intelligence ecosystem to fend off rising competition.

The chip designer built its reputation on speedy chips, but it also offers a range of its own AI models, from physics simulations to self-driving vehicles, as open-source software that researchers and companies can use.

Its proprietary CUDA software, a standard among most developers, is a major selling point for its chips, making software key to maintaining its dominance in the AI industry.

SchedMD provides software that helps schedule large computing jobs that can occupy a big share of a data centre's server capacity.

Its technology, called Slurm, is open source, meaning developers and firms can access it for free, while the company sells engineering and maintenance support.

Financial terms of the deal were not disclosed.

Nvidia said it would continue to distribute SchedMD's software on an open-source basis.

"Slurm, which is supported on the latest Nvidia hardware, is also part of the critical infrastructure needed for generative AI, used by foundation model developers and AI builders to manage model training and inference needs," Nvidia said in a blog post.

Earlier, Nvidia unveiled a new family of open-source AI models that it says will be faster, cheaper and smarter than its previous offerings, as it faces a growing wave of rival open-source models from Chinese AI labs.

SchedMD was founded in 2010 by Slurm software developers Morris "Moe" Jette and Danny Auble in Livermore, California, and the company currently employs 40 people, according to its website.

Its customers include cloud infrastructure firm CoreWeave and the Barcelona Supercomputing Center, among others.