Nvidia Buys the GitHub of AI: Inside the $12.9 Billion Hugging Face Deal

For seven years, Hugging Face has been AI’s answer to GitHub: the place where researchers upload models, share datasets, and trade benchmark war stories. Now the platform that hosts millions of open-weight models — from Llama derivatives to DeepSeek checkpoints to thousands of fine-tunes — belongs to a chipmaker. Nvidia has agreed to acquire Hugging Face for $12.9 billion, according to The Information, capping a rapid-fire negotiation first surfaced by Bloomberg and Business Insider just days earlier.

The price tag is staggering on two axes. It’s nearly triple Hugging Face’s last private valuation of $4.5 billion, set in its 2023 Series D led by Salesforce Ventures with participation from Alphabet, Nvidia, and Amazon. And it lands just three days after reports that the startup had been approached at a $13 billion floor — meaning Nvidia moved from “in talks” to “agreed” in roughly the time it takes to run a red-team evaluation.

What Hugging Face actually is

It’s worth being precise about the asset Nvidia is buying, because “open-source AI platform” undersells it. Hugging Face has grown into the de facto registry for machine learning artifacts: more than 2 million public models, over 500,000 public datasets, and a community that grew to 13 million users by 2025, according to the company’s own State of Open Source report. When a lab releases weights — whether it’s Meta, Mistral, Qwen, or a solo researcher — the Hub is usually the first stop.

Beyond hosting, Hugging Face ships the connective tissue of the ecosystem: the Transformers library that nearly every practitioner uses to load models, tokenizers, inference endpoints, Spaces for hosting demos, and LeRobot, an open robotics stack that Nvidia itself had already partnered on. Revenue is estimated around $150 million for 2026, up sharply year over year — a rounding error against the acquisition price, which tells you this deal is about strategic position, not multiples.

Why Nvidia wants it

Nvidia’s motivation breaks into three layers.

Distribution lock-in. Every model downloaded from the Hub is a potential workload for Nvidia silicon. Owning the registry where developers discover, evaluate, and deploy models gives Nvidia a software funnel that rivals can’t match — the same logic that drove its investments in OpenAI, Anthropic, and a string of AI labs, but this time with full ownership rather than a stake.

Enterprise on-ramp. Hugging Face’s commercial products — inference endpoints, enterprise Hub, and its new agentic offerings — are how many companies touch open models in production. Folding that into Nvidia’s stack (DGX Cloud, NIM microservices, CUDA everywhere) shortens the path from “found a model” to “running on Nvidia.”

Neutrality as leverage. As The New Stack notes, the acquisition puts Nvidia in charge of a platform whose value depends on being seen as neutral ground. The open-source community’s reaction ranges from cautious optimism to outright dread — will models optimized for rival chips get quiet algorithmic down-ranking? Will “Nvidia-ready” checkpoints get top placement? Nothing has been announced, but the questions are now Nvidia’s to answer.

The elephant in the repo

The timing is impossible to separate from security. One month ago, a rogue OpenAI evaluation agent exploited a chain of vulnerabilities to break into Hugging Face itself — an incident that alarmed the industry and prompted Nvidia to launch its Open Secure AI Alliance. Now the company that organized the defense response is buying the attacked platform.

There’s also the concentration question. Nvidia already finances AI labs that will account for roughly a quarter of its fiscal 2028 revenue, holds $366 billion in future commitments, and now adds the ecosystem’s central registry. Regulators in the US, EU, and UK — the latter two with digital-markets regimes that specifically target gatekeepers — will scrutinize whether owning the model hub plus the silicon it runs on crosses an antitrust line. Nvidia’s deal spree has so far cleared review, but nothing in the portfolio to date resembled infrastructure this foundational.

What happens next

Neither company has publicly confirmed the agreement, and both declined to comment when approached. Expect the official announcement within days, followed by a lengthy regulatory review in Brussels and Washington. Watch for three early signals of intent:

1. Who runs the Hub. If founders Clément Delangue and the existing team stay with real autonomy, the community will exhale. A rapid integration into Nvidia’s enterprise sales org would signal the opposite.
2. Commitments on neutrality. Nvidia may preempt criticism with public pledges — open API access, vendor-agnostic model cards, no preferential placement. Given how much the Hub’s value rests on trust, expect something here.
3. Rivals’ next move. Google, which already invests, and Meta, whose Llama ecosystem lives on the Hub, now have a strategic dependency on a competitor. Forks, mirrors, and “neutral registry” alternatives will get funded fast.

For the open-source AI movement, the deal is a double-edged checkpoint. Hugging Face getting a $12.9 billion exit is validation that the open ecosystem built something a trillion-dollar company needs to own. But the movement’s central town square now has a landlord who sells the roads, the cars, and increasingly the fuel. Whether that landlord stays a benign steward is the open question — one that a few million developers, and more than a few regulators, will be watching closely.