Nvidia is dropping $12.9 billion to acquire Hugging Face, the open-source AI platform that’s been the go-to hub for developers sharing models for the past decade. The Information broke the story, and the numbers are striking: Hugging Face pulls in roughly $150 million in annual revenue and sits ‘close to profitability,’ according to CEO Clem Delangue. That means Nvidia is paying about 80 times annual revenue — a price that only makes sense if you’re buying something strategic, not just a business.
Here’s the context that makes this deal interesting. Nvidia already committed $26 billion over five years to open-source AI development. Meanwhile, the big closed labs are actively trying to wean themselves off Nvidia hardware. OpenAI has inked deals with Broadcom for custom chips, signed a billion-dollar agreement with Cerebras, and locked in access to AMD silicon. Anthropic is working on its own chip. Google’s been at it for years. The message from the frontier labs is clear: they don’t want to be held hostage by Jensen Huang’s supply chain.
So where does Hugging Face fit? Think of it as potential distribution. Nvidia could scale up Hugging Face’s cloud business and position it as the OpenRouter for open-weight models — a place where developers spin up whatever open model they want, running on Nvidia hardware. Hugging Face already has enormous developer mindshare. If Nvidia owns the platform where millions of developers interact with open models, that’s a built-in demand engine for its chips.
It’s a defensive move wrapped in an open-source flag. The closed labs are building their own silicon moats, so Nvidia is doubling down on the open ecosystem instead. Whether developers buy it — or whether this just becomes another corporate acquisition that slowly strangles the community it acquired — is the real question nobody can answer yet.
