Nvidia Buys Model Hub Hugging Face: Its Openness Promise Is Unwritten

On Thursday, September 3, Nvidia agreed to acquire Hugging Face for $12.93 billion.

Hugging Face makes no chips and sells no cloud. It is a catalogue of open-weight AI models: more than 3 million models, 500,000 datasets and 1 million applications, used by over 200,000 companies and 18 million developers.

Against that sits an annualized run rate of roughly $150 million, about 0.16% of a single Nvidia quarter; Nvidia reported $96.2 billion of revenue in the quarter ended July 26.

Why pay $12.93 billion for a company generating $150 million in revenue?

What the Money Actually Buys

Founded in 2016 as a chat application, Hugging Face arrived at its present business through its open-source libraries.

In 2023 it raised $235 million in a round led by Salesforce at a $4.5 billion valuation. Nvidia was in that round.

A 2.9x increase in three years. What closed the valuation gap was not the product but the position: Hugging Face became the default front door for open models.

Per InfoWorld, roughly $11.9 billion of the price is the acquisition itself and the remaining $1 billion or so equity-based retention for about 750 employees.

The Mechanism: 86 Years or 15 Days

Which revenue stream pays the price back? Two calculations, very far apart.

The asset does not need to pay for itself out of its own revenue.

The moment a developer selects a model is the moment the hardware it runs on is decided. Nvidia paid for that moment.

$12.93 billion is 21.7% of the $59.7 billion in net income Nvidia reported last quarter. Per developer, the price works out to $718.

What the Promise Covers, and What It Does Not

Nvidia committed to keeping the platform open. Huang's sentence is specific:

"NVIDIA compute will not be required to build on or deploy through Hugging Face."

What was said is that Nvidia compute will not be required.

Is there an instrument that makes that binding? No competition-authority decision, no consent agreement — a blog post and a chief executive's statement.

Sanchit Vir Gogia of Greyhound Research draws the distinction:

"Nvidia's openness commitment is precise where it is cheap, and silent where it is expensive."

The commitment governs requirement, not ranking.

Where a model appears in search, which runtime is suggested by default, which hardware the one-click configuration targets — none of it falls inside the promise.

Nithya Ruff, chair of the Linux Foundation board: neutrality is a discipline chosen again and again, not a promise made once.

This Time There Will Be a Filing

The least-covered aspect is the deal's legal form: Nvidia has done three large transactions in nine months and structured none of them as acquisitions.

In Groq it did not buy the company: it licensed the inference-chip technology non-exclusively and hired key staff, the chief executive included.

Poolside was a $6 billion non-exclusive license plus a $1 billion minority investment at a $12 billion pre-money valuation, with 109 employees moving across.

The common result: staying outside the merger-notification threshold.

Hart-Scott-Rodino requires advance notification above a certain size. The FTC's 2026 threshold, announced in January, is $133.9 million; a license plus a minority stake does not trip it.

The three deals total roughly $28 billion. None was notified.

Senators Elizabeth Warren and Richard Blumenthal wrote to Jensen Huang on March 23, 2026. Groq was an acquisition in all but name, they argued, and Nvidia holds roughly 90% of the GPU market.

FTC Chair Andrew Ferguson said in January that the agency was examining whether such structures were built to escape notification.

Hugging Face sits outside that sequence: a direct $12.93 billion acquisition is far above the threshold.

Justin Boitano of Nvidia argues open-source platforms are by definition deconcentrating. The opposing case starts from the same place: whoever owns that layer sets its defaults.

The Same Weeks, a Second Layer

In mid-August Stripe acquired OpenRouter — not a model developer either, but a gateway routing requests across providers.

Reported prices vary: Bloomberg and TechCrunch put it above $7 billion, a figure attributed to Reuters a little above $8 billion.

The models stay open. The infrastructure around them does not have to.

What the Market Did

On September 3 the S&P 500 rose 0.48% and the Nasdaq 0.84%. The decisive moves were elsewhere: Snowflake rose 21.9% after earnings, Broadcom fell 6.4% on weak guidance.

For a company with $96.2 billion of quarterly revenue guiding to $108 billion, a $12.93 billion acquisition will not show up in the income statement, nor against its market capitalization.

What Is Left

Closing is expected in the first half of 2027. That calendar includes the US waiting period and any second request, plus Phase I and possible Phase II in Europe.

Then the defaults: the ordering of model cards, the suggested runtime, which hardware comes first in prepared configurations.

Then enterprise behaviour: whether firms mirror models into their own registries, separating discovery from custody.

This article draws on Nvidia's announcement on its corporate blog, the company's results release for the quarter ended July 26 as filed with the SEC, InfoWorld's reporting on the deal structure, sector analysis from SiliconANGLE and VentureBeat, the letter text published by Senator Warren's office, and the FTC's announcement of the 2026 Hart-Scott-Rodino thresholds. Hugging Face's revenue run rate comes from Dealroom; the split between the $11.9 billion price and roughly $1 billion in retention equity comes from InfoWorld; the price range for the OpenRouter transaction comes from Bloomberg, TechCrunch and reporting attributed to Reuters. The deal has not closed and remains subject to regulatory review.

Correction (September 13, 2026): prose tightened for rhythm; figures and sources unchanged.