Nvidia bets $12.9B that owning Hugging Face won’t kill open-source AI

Hugging Face is the infrastructure layer that most AI development quietly depends on. Its model repository hosts hundreds of thousands of AI models — from lightweight tools to frontier-scale architectures — available to any developer without a corporate agreement or a cloud subscription. More than ten million practitioners use it. In a field where compute access and proprietary model APIs can cost hundreds of thousands of dollars a month, it is where you go to build without asking permission.

Nvidia, according to reporting by Bloomberg, TechCrunch and Fortune, has reportedly reached a deal to acquire the company for $12.9 billion — nearly three times Hugging Face’s last known private valuation. If completed, the acquisition would give the world’s dominant AI chip company ownership of the platform most of the field treats as neutral ground.

Why Nvidia needs this

Nvidia’s GPU business depends on a straightforward structural fact: more AI built means more compute purchased. That logic is under pressure. Google, Amazon, Microsoft and Meta have all invested heavily in custom AI accelerator chips — the specific market Nvidia has dominated for five years. If major cloud providers migrate AI workloads onto their own silicon, Nvidia loses the revenue those workloads generate.

Owning Hugging Face changes the equation. Hugging Face hosts not just models but the full training-and-deployment pipeline, and running that pipeline at any real scale requires GPU compute. Under Nvidia, those requirements could be steered toward Nvidia infrastructure. The company previously operated DGX Cloud, a cloud AI service it pulled back from — Hugging Face gives it a rebuilt path into that market, with an existing user base of ten million developers.

What is still unsettled

No signed agreement was in place as of the reporting date, according to Business Insider. Deals of this size regularly get announced before they are concluded, and reporting from multiple outlets is consistent with an active negotiation, not a completed transaction. Neither Nvidia nor Hugging Face confirmed the reports at the time of publication.

There is also the structural question of what happens to Hugging Face’s reason for existing. CEO Clem Delangue has built the company on an explicit position: open models are a public good, and concentrated control over them is a risk. Hugging Face inside Nvidia would be controlled by a company whose financial interest is in selling more GPU compute — including to developers who currently run models locally, on hardware Nvidia does not sell, to avoid exactly that dependency. Whether the platform can maintain genuine neutrality inside that ownership structure is a question the deal does not resolve.

Hugging Face was founded in 2016 and pivoted from a chatbot service to become the primary open-source AI model repository. A 2023 Series D round valued the company at $4.5 billion and brought in investment from Google, Amazon, Salesforce and Nvidia. Nvidia had separately offered to lead a follow-on round at a $7 billion valuation; Hugging Face declined. The reported $12.9 billion acquisition price represents nearly twice what Nvidia’s own investment team put on the company less than two years ago.

If completed, the deal would require regulatory review in the United States and Europe — a process that typically takes six months or more for acquisitions of this scale. Annual revenue at Hugging Face was approximately $150 million at the time of reporting, placing the acquisition multiple at roughly 85 times trailing revenue — a price that reflects what the platform is positioned to become, not what it currently generates.

Tags: nvidia, Hugging Face, open-source AI, AI acquisition, Clem Delangue