Nvidia's $12.9 Billion Hugging Face Acquisition, Explained

Two days after reports said Hugging Face was quietly testing acquisition interest without naming a buyer, the buyer got a name. On August 26 and 27, 2026, The Information and Business Insider reported that Nvidia has agreed, or is very close to agreeing, to acquire Hugging Face, the platform where the open AI world hosts, versions and distributes its models, for roughly $12.9 billion. As of this writing, neither company has officially confirmed the deal, and Business Insider's own account describes talks that had not yet produced a signed agreement and could still fall apart. But the reporting is specific enough, and consistent enough across outlets, that the deal is worth taking seriously as reported. This is a distinct story from the earlier sale rumor: the price is now attached to a name, and that name is the company that already dominates the layer underneath everything Hugging Face hosts.

NVIDIA

OpenAI

What the reports actually say

Hugging Face, founded in 2016, has grown into the default place developers go to publish, download and version open AI models, datasets and the libraries used to load them, alongside a cloud inference business built on top. According to the reporting, Nvidia's price values the company at about $12.9 billion, a figure The Information's sourcing describes as agreed, while Business Insider frames the same talks as still short of a signed contract. Hugging Face's annualized revenue has been reported at roughly $150 million as of late August 2026, up from about $100 million just two months earlier, with chief executive Clement Delangue describing the company as approaching profitability. At the reported price, that puts the multiple at more than 80 times run-rate revenue, a number that only makes sense if the buyer is paying for a strategic position rather than current cash flow.

That position has a history with Nvidia specifically. Nvidia was already an investor, having put $235 million into Hugging Face's 2023 Series D at a $4.5 billion valuation. Late in 2025, Nvidia reportedly proposed investing $500 million more at a $7 billion valuation, an offer Hugging Face turned down. Less than a year later, the number on the table nearly doubled again, and this time it is reportedly for the whole company rather than a stake in it.

Nvidia's price for Hugging Face nearly tripled in three years

Disclosed and reported valuation marks. The 2026 figure is a reported price, not a confirmed, signed number.

2023 Series D · Nvidia invested $235M in the round$4.5B

Late 2025 offer · Nvidia proposed $500M; Hugging Face declined$7B

Aug 2026 reported deal · Full acquisition, per The Information$12.9B

Hugging Face's annualized revenue was reported at roughly $150 million in August 2026, up from about $100 million two months earlier, implying the reported $12.9 billion price is on the order of 80 times run-rate revenue.

Why a chip company wants to own the hub

The strategic logic reported around the deal centers on a threat to Nvidia's core business that has nothing to do with model hosting on its face: several of Nvidia's largest customers, including OpenAI, are developing their own custom AI chips, and other hyperscalers have been doing the same for years. Every dollar spent on in-house or rival silicon is a dollar that does not flow to Nvidia. Keeping the open model ecosystem healthy is one of the more durable defenses against that, because open weights that run well on Nvidia hardware give developers less reason to optimize exclusively around a narrower set of proprietary chips. Owning the place where that ecosystem lives, gets discovered and gets deployed is a more direct lever on that outcome than simply investing in the labs that publish there.

There is a second, more conventional motive layered on top. Hugging Face's business already includes cloud inference infrastructure, giving Nvidia a route back into a layer closer to end customers than selling chips to cloud providers, plus a way to put idle capacity from its own compute commitments to work. Jensen Huang has been publicly explicit about why Nvidia cares about open models at all, arguing that "open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty." Owning the primary distribution channel for those models backs that stated position with control, not just capital.

Hugging Face's value to the open source community depends on its perceived neutrality, and a chip company owning it is a real tension with that positioning.
Reported concern raised in coverage of the deal

The neutrality problem the deal doesn't solve

Set the strategic logic next to what makes Hugging Face valuable in the first place, and the tension is immediate. Developers, and the labs themselves, from OpenAI to Meta to Mistral to DeepSeek, publish and pull models on Hugging Face partly because it belongs to none of them. That perceived neutrality is the asset, and it is also the one thing a strategic buyer cannot purchase without at least partially consuming it. Coverage of the reported deal has raised exactly that concern: a chip company sitting at the center of open model distribution changes the incentives of the party that controls discovery, hosting terms and default tooling, even if nothing about the product changes on day one.

Where Hugging Face sits, and what ownership would change

Hugging Face is the distribution layer between the labs that train open models and the developers who deploy them. A reported deal would put that layer inside a chip supplier for the first time.

Model publishers

OpenAI, Meta, Mistral AI, DeepSeek, Qwen, Google and independent labs publish open weights, datasets and libraries.

↓ publish to

The hub (reported acquisition target)

Hugging Face

Hosts and versions open models and datasets, provides the libraries most teams use to load them, and runs cloud inference infrastructure. Reported price: $12.9 billion.

↓ distribute to

Developers and enterprises

Teams pull weights, push fine-tuned variants, and build products on top, largely assuming the hub is neutral among the labs that publish there.

↑ would sit inside

The reported buyer

Nvidia

Already the dominant supplier of the chips every layer above runs on. Owning the hub would extend that position into software distribution and developer relationships.

Diagram simplifies a many-to-many relationship; not every model publisher or user of Hugging Face is shown.

NVIDIA founder and CEO Jensen Huang speaking to students at Stanford University in April 2026 — NVIDIA founder and CEO Jensen Huang speaking at Stanford University in April 2026. Huang has publicly framed open models as good for safety, competition and sovereignty, a stance that lines up with why Nvidia is reported to want to own their primary distribution hub. Photo via Wikimedia Commons, CC BY-SA 4.0.

None of that means the deal is bad for the ecosystem by default. A financially secure Hugging Face, backed by a company with every incentive to keep open weights competitive with closed alternatives, could invest more in the hosting and tooling developers already rely on. But it is a different kind of ownership than a diversified investor base or an independent buyer would have provided, and it puts a decision that used to sit with a standalone company inside the roadmap of the world's largest AI chip supplier.

What it means for teams building on open models

For any team whose workflow runs through Hugging Face today, the practical takeaway is not to panic about a deal that has not been officially confirmed, but to notice how much of the open model supply chain now runs through a small number of chokepoints, ownership included. That is the same argument for staying model-agnostic at the application layer that shows up whenever infrastructure ownership shifts underneath a widely used product. Tools built to route across providers rather than commit to one, the way Metir AI treats the underlying model as a swappable choice rather than a fixed dependency, are one practical hedge against exactly this kind of concentration, whichever company ends up owning which layer.

What to watch next

The simplest open question is whether this becomes an official, signed deal at all, given that Business Insider's own sourcing still describes it as unresolved. If it does close, the next things worth watching are whether Hugging Face's hosting priorities shift for models that compete with Nvidia's own interests, whether other chip vendors or cloud providers respond by backing a competing hub, and whether regulators in the US or EU take an interest in a deal that hands a dominant hardware supplier control of software infrastructure much of the AI industry depends on. Nvidia has spent the past several years buying stakes in, and in some cases acquiring outright, pieces of the AI stack around its chips. Hugging Face, if the reported price holds, would be the biggest and most structurally significant of those moves yet.

Sources:

Image credits

Header image: NVIDIA's Endeavor headquarters building in Santa Clara, California, by Coolcaesar via Wikimedia Commons, licensed under CC BY-SA 4.0. In-body photograph of NVIDIA founder and CEO Jensen Huang speaking at Stanford University on April 30, 2026, by Anderseidesvik via Wikimedia Commons, licensed under CC BY-SA 4.0.