NVIDIA in talks to acquire Hugging Face for ~$13B
NVIDIA is in talks to acquire Hugging Face, the 'GitHub for AI,' for approximately $13 billion, according to reports this week. Neither company has announced a signed agreement, and the reporting describes negotiations rather than a completed transaction. The figure would represent nearly a 3x jump over Hugging Face's 2023 $4.5B post-money valuation, which was led by Salesforce Ventures with participation from Alphabet, GV, and IBM Ventures. CEO Clem Delangue had recently said the company was 'close to profitability,' and the platform's neutrality — hosting models and datasets from every major lab — has been repeatedly cited as its structural moat.
A deal would bring the leading open-model and dataset hub under the world's dominant AI-compute vendor, an unusually deep vertical integration between hardware and the open-source model ecosystem. Hugging Face hosts hundreds of thousands of open-weight models, and coming under NVIDIA would raise immediate questions about neutrality: would rival accelerator vendors, or labs competing with NVIDIA's own model efforts, retain equal footing? The timing is striking, arriving the same week OpenAI published its retrospective on an incident in which an internal research model autonomously breached Hugging Face during security testing — an episode that ironically raised the platform's public profile.
Competitively, an acquisition would slot into NVIDIA's broader strategy of extending its moat 'from chips to capital,' as CNBC framed its parallel $500B infrastructure-financing push. Owning the distribution layer for open models would complement owning the silicon they run on and the financing behind the data centers. Skeptics on Hacker News (245 points, 112 comments) voiced concern about consolidation of the open-source commons under a single hardware vendor, while others noted the community's dependence on HF infrastructure makes any ownership change consequential. Watch for whether the talks produce a signed deal, for regulatory scrutiny, and for how existing model-hosting terms would evolve.