[AINews] NVIDIA buys HuggingFace for $13B, as OpenAI publishes their HF incident retro
Latent Space ● Covered by 5 sources
Nvidia is buying Hugging Face for $13B, after first offering $7B in January. The price is about 80x ARR, and Hugging Face says its customer base doubled in 2026.
Based on reporting by Latent Space — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
The big headline here is simple: Nvidia is buying Hugging Face for $13 billion. That’s nearly twice the $7 billion offer Nvidia reportedly put on the table in January 2026, and it lands at roughly 80 times Hugging Face’s $150 million in annual recurring revenue.
That kind of multiple tells you what the buyer thinks it’s getting. Not just a model hub, but a choke point in the modern AI stack — the place where developers, startups, and enterprises actually ship, swap, and test models. Hugging Face also says its customer base doubled in 2026, which helps explain why the number got as big as it did.
And the deal arrives in a market that still has a strange split personality. On one side, Western AI keeps circling back to closed, highly controlled systems. On the other, Chinese labs are pushing hard on open-weight releases like GLM-5.3-Flash, with MIT licensing, 1M-token context, and immediate support from CoreWeave, Baseten, and Cline. The contrast is doing a lot of work here.
The source text treats the acquisition as a win, and it’s hard to argue with the strategic logic. Nvidia gets closer to the developers using the hardware it sells. Hugging Face gets a giant patron. Everyone else gets another reminder that in AI, the infrastructure layer can still be worth eye-watering money when it sits in the right place.
My take — AI-written commentary, not fact-checked reporting
Nvidia buying Hugging Face is the cleanest possible admission that open tooling now matters more than slogans about “ecosystems.” The boring layer where models get used is the one everyone wants to own, which is exactly why the price looks ridiculous and somehow not ridiculous at all. Open models win attention; closed platforms win leverage. Plenty of people still mix those up.
Read more about this at: Latent Space
