Nvidia is prepping an open-source AI agent platform called NemoClaw, pitched as a rival to OpenClaw.
Wired says it will let enterprise software companies send AI agents off to do chores for their own staff, without needing a babysitter. Companies will be able to access the platform regardless of whether their products run on Nvidia’s chips.
Nvidia is heading into its annual developer conference in San Jose next week and looking for extra noise. Ahead of that, it has been chatting to Salesforce, Cisco, Google, Adobe and CrowdStrike about partnerships for the agent platform.
Because the platform is open source, any partner deal sounds less like a fat cheque and more like early access and some code contributions. Nvidia is expected to bundle security and privacy tools into the project.
NemoClaw looks like Nvidia trying to charm the software crowd with tighter controls for agents that could rummage through company systems. It fits with Nvidia leaning harder into open-source AI models while big AI labs crank out custom chips to dodge dependencies.
That is awkward for a company whose software strategy has been welded to CUDA, a famously proprietary setup that keeps developers penned into Nvidia GPUs. NemoClaw reads like a polite attempt to widen the funnel without giving up the “moat” that made Nvidia so rich.
