Nvidia’s $12.9 Billion Hugging Face Deal, Explained
On this page
Nvidia has reportedly agreed to buy Hugging Face, the open-source hub often called “the GitHub of AI,” for about $12.9 billion. The figure first surfaced in The Information on August 26, 2026, was echoed by Bloomberg and Business Insider, and values a company with roughly $150 million in annual revenue at nearly three times its last public valuation. As of this writing neither Nvidia nor Hugging Face has confirmed anything on the record, so treat the price and the “signed” language as reporting rather than fact. But the direction of travel is hard to miss: the company that sells most of the world’s AI chips wants to own the place where most of the world’s AI models are shared.
I have been downloading models and datasets from Hugging Face for years, and my first reaction was less about the money and more about neutrality. The whole reason that site works is that it does not care what hardware you run. A deal like this puts that assumption in play.
What is actually being reported
Business Insider broke the story of “serious talks” over the weekend of August 22. The Information followed on August 26 with a single source describing an agreement at $12.9 billion, and Bloomberg reported the same day that the two sides had discussed a purchase. TechCrunch and Fortune both noted they could not independently verify a signed deal, and that Business Insider’s sources said the talks “had not yet produced a signed agreement and could still fall apart.”
One detail worth flagging: Nvidia has a track record of quickly pushing back on reports it considers wrong, and it stayed quiet here. That silence is not confirmation, but several reporters read it as telling.
Quick reference: the deal at a glance
| Item | Detail |
|---|---|
| Reported price | ~$12.9 billion (Business Insider: “over $13 billion”) |
| First reported | Business Insider, weekend of Aug 22, 2026; The Information, Aug 26 |
| Official confirmation | None from Nvidia or Hugging Face as of Aug 28, 2026 |
| Hugging Face founded | 2016 (started as a chatbot app, pivoted to model hosting) |
| Last valuation | $4.5 billion (2023, $235M round led by Salesforce Ventures) |
| Annual revenue | ~$150 million, reportedly near profitability |
| Platform scale | ~13 million users, 2M+ public models, 500,000+ datasets |
| Prior Nvidia offer | $500M investment at ~$7B valuation, turned down in late 2025 |
Why Nvidia would pay this much
Hugging Face does not make a lot of money. It makes a lot of habits. When a developer starts a new project, the first stop is usually a model card on Hugging Face, a dataset, a quick test in a hosted Space, maybe an inference endpoint. That is the top of the funnel for the entire AI industry, and Nvidia has spent years trying to sit at the top of every funnel it can find.
The strategic logic is the same one CEO Jensen Huang repeats in public: wider access to open models drives more AI adoption, and more adoption drives demand for chips and data centers. Nvidia already ships the CUDA libraries, the TensorRT optimizer, and the NIM microservice packaging that most of those models eventually run on. Owning the catalog lets it wire that stack in by default, so the smoothest path from “found a model” to “running in production” runs on Nvidia silicon.
There is also a cloud angle. Hugging Face already rents compute through Inference Endpoints and Spaces. Nvidia has wanted back into the business of selling hosted compute directly to developers, and buying an established platform is faster than building one. Its recent infrastructure moves, like the $105 billion Ohio data center partnership with OpenAI, show how aggressively it is building capacity that needs to be filled.
The neutrality problem
Hugging Face’s board reads like a truce between rivals. Its investors include Google, Amazon, AMD, Intel, Qualcomm, IBM, ServiceNow, Salesforce, and Nvidia itself. Its engineering partnerships span AMD’s ROCm software, AWS’s custom Trainium and Inferentia chips, and Google’s TPUs. The platform is useful precisely because it treats all of that hardware as a first-class citizen.
Hugging Face understood this well enough to turn down half a billion dollars from Nvidia last year, reportedly because it did not want one dominant backer steering its decisions. An outright acquisition is that concern taken to its logical end. If the hub is owned by a chip company with a direct financial interest in developer lock-in, every competing hardware vendor has to ask whether its models will still get equal treatment, equal benchmarking, and equal placement.
Expect regulators to ask the same question. Nvidia’s proposed $40 billion purchase of Arm collapsed in 2022 under antitrust pressure on three continents, and that deal did not touch a resource this many competitors depend on. Analysts already expect scrutiny in the US, the EU, and possibly the UK. A deal announced today could take a year or more to clear, if it clears at all.
What changes for people who use the platform
In the short term, probably very little. Acquisitions of this size move slowly, and Hugging Face’s open-source libraries, transformers, diffusers, datasets, and the rest, are permissively licensed and widely mirrored. If the platform’s direction shifted hard, the community has the tools to fork what it needs.
The things I would watch over the next year: whether model optimization and “recommended” deployment paths quietly default to Nvidia hardware, whether support for AMD, Google, and AWS accelerators keeps pace, whether pricing on hosted inference changes, and whether any senior Hugging Face staff leave. The company’s culture has been loudly pro-open and vendor-agnostic. Keeping that intact inside Nvidia would be the real test, and the honest answer is that nobody knows yet.
It also lands during a striking week for Nvidia, which reported another record quarter and guided to roughly 70 percent revenue growth for the year ahead on the back of AI infrastructure demand. When a company is growing that fast, a $12.9 billion acquisition is a rounding error against its cash pile, which is part of why the story is so plausible even without confirmation.
Frequently asked questions
Is the Nvidia–Hugging Face acquisition confirmed?
No. As of August 28, 2026 it is based on reporting from The Information, Bloomberg, and Business Insider citing unnamed sources. Neither company has issued a statement, and at least one outlet’s sources said a deal could still fall through.
How much is Nvidia reportedly paying?
About $12.9 billion, with Business Insider putting the valuation above $13 billion. That is roughly triple Hugging Face’s $4.5 billion valuation from its 2023 funding round.
Why does a company with $150 million in revenue cost $12.9 billion?
The value is in position, not profit. Hugging Face is the default starting point for millions of AI developers, hosting more than two million models. Owning that catalog gives Nvidia influence over how models are discovered, optimized, and deployed, which ultimately feeds demand for its chips.
Could regulators block the deal?
It is possible. Nvidia’s $40 billion Arm acquisition failed in 2022 under global antitrust pressure, and this deal raises similar concerns because rival chipmakers and cloud providers rely on Hugging Face staying neutral. Analysts expect close review in multiple jurisdictions.
Should I stop using Hugging Face?
There is no immediate reason to. The core libraries are open source and mirrored widely, and nothing has changed operationally. It is worth keeping local backups of models and datasets you depend on, which was good practice before this news anyway.
The bottom line
If the reporting holds, Nvidia is buying the connective tissue of open-source AI. That is a smart move for Nvidia and a genuinely uncomfortable one for everyone who relied on that tissue being neutral. The next signal to watch for is simple: an official statement, or a pointed non-denial, from either company. Until then, this is the biggest maybe in tech. Hugging Face has sat at the center of industry drama before, including the incident where OpenAI test models breached its infrastructure, and you can read TechCrunch’s original report on the acquisition talks for the primary reporting.
Jared is a tech journalist covering product launches, industry news, and the culture around technology. He has been reporting on the consumer tech beat for more than eight years.
