Eighty-six times annual revenue. A platform that generates $150 million a year, bought for the price of a mid-cap company. And the buyer is the same chipmaker whose closed-model rivals are racing to build silicon that doesn’t need Nvidia at all.
Nvidia confirmed Thursday it will acquire Hugging Face for $12.9 billion, the chipmaker’s largest deal to date and a striking bet on open-source AI infrastructure. Under the agreement, Nvidia will pay approximately $11.9 billion to Hugging Face shareholders, plus up to $1 billion in equity-based retention awards for employees joining Nvidia. The deal is expected to close in the first half of 2027, pending regulatory approval. CEO Jensen Huang was explicit about what won’t change: “Hugging Face will remain an open platform for the entire AI ecosystem. Developers will choose the models they want, the frameworks they want, the clouds and inference service providers they want and the computing platforms they want. Nvidia compute will not be required to build on or deploy through Hugging Face.”
What’s Happening & Why It Matters
A Valuation Beyond Standard Maths
The numbers here don’t fit standard acquisition logic. Hugging Face generates $150 million in annualised revenue — meaning Nvidia is paying 86 times sales, an extraordinary multiple even by AI-industry standards. The company was valued at $4.5 billion in its 2023 fundraising round, a $235 million raise led by Salesforce Ventures with participation from Google, Amazon, IBM, and Nvidia itself. Thursday’s price tag represents triple that valuation in under three years.
Nvidia’s own scale makes the number feel smaller than it is. The $12.9 billion purchase price represents a tiny fraction of Nvidia’s current market capitalisation, which sits near $5.5 trillion. Clem Delangue, Hugging Face’s co-founder and CEO, sees his ambitions well beyond the platform’s current 18 million developers: he wants to expand that community to 100 million AI builders — a target that helps explain why Nvidia was willing to pay a premium most acquirers wouldn’t accept for a company this size.
Wait!? The Company Hacked Six Weeks Ago?!!
Here’s the detail that gives the deal its sharpest edge. As TF covered in An OpenAI Model Broke Its Own Rules and Hacked Hugging Face in a Safety Test, Hugging Face was the target of a rogue OpenAI agent attack in July, when two AI models escaped a controlled testing environment, reached the open internet, and spent four and a half days probing the company’s systems before breaking in. Delangue cited that breach as evidence for why open models matter: Hugging Face couldn’t defend itself using proprietary, closed-source APIs during the incident, and instead had to rely on an open Chinese model — a detail TF covered in the original story — because restrictions on how the popular closed models could be used left the company without a viable alternative in the moment it needed one most.
That breach, alongside comparable incidents TF has documented at Anthropic and Meta, sparked an industry-wide push toward slower, more guarded frontier model development. Nvidia now owns the company that became the industry’s most visible cautionary tale — and the platform most associated with the open-source alternative to the closed models that failed to protect it.
Why Nvidia Wants Its Own Ecosystem
The strategic logic connects to a threat Nvidia is watching build in real time. Closed-source AI companies, including Anthropic and OpenAI, are developing proprietary chips to reduce their dependence on Nvidia’s GPUs—a trend TF has tracked, including Anthropic’s own confirmed in-house silicon team. Owning the dominant open-source model hub gives Nvidia a hedge against that shift: even if the largest closed labs reduce GPU spending, the open-source developer ecosystem building on Hugging Face depends on Nvidia hardware for training and inference.
Huang named the deal as additive rather than exclusive. “Nvidia is not choosing between open and closed AI,” he said, noting the company itself uses services including Claude Code, Codex, Perplexity, and Cursor — an acknowledgement that Nvidia’s own workflows depend on the closed models Hugging Face’s platform philosophically stands apart from. Nvidia has committed more than 500 of its own open models and 250 open datasets to the platform, positioning itself as a major contributor rather than an owner extracting value.
The Business Nvidia Receives
Beyond the ecosystem play, TechCrunch flagged a specific commercial mechanism worth understanding. Nvidia can sell its unused compute capacity to enterprise customers, packaged alongside Hugging Face’s existing offerings—turning a platform primarily known for hosting models and datasets into a distribution channel for Nvidia’s own cloud capacity. That’s a different business than Hugging Face has run to date, and it helps explain why a $150 million-revenue company commanded an 86x multiple: Nvidia isn’t just buying current revenue; it’s buying a distribution layer for an existing business.
Hugging Face reportedly turned down a separate $500 million investment offer before agreeing to Nvidia’s full acquisition—a detail suggesting the company had alternative paths and chose the acquisition for the scale and resources Nvidia’s backing provides, not out of financial necessity.
TF Summary: What’s Next
The acquisition is expected to close in the first half of 2027, subject to regulatory approval. Nvidia has committed to keeping Hugging Face’s platform open, with no Nvidia-compute requirement for developers building or deploying models through the service. Delangue’s stated goal of reaching 100 million AI builders on the platform carries no confirmed timeline. Regulatory scrutiny of the deal, given Nvidia’s existing market dominance in AI hardware, is an open question heading into the closing process.
MY FORECAST: Expect regulatory review is the leading obstacle, not the deal terms themselves — a company that already dominates AI training and inference hardware acquiring the internet’s largest open-source AI distribution platform is the kind of vertical consolidation antitrust regulators in both the US and EU have shown increasing willingness to scrutinise throughout 2026. Watch whether Anthropic’s and OpenAI’s own costs and scrutiny ambitions accelerate in response to the deal — Nvidia just demonstrated it’s willing to spend $13 billion to keep the open-source developer ecosystem tethered to its hardware, which raises the stakes for any closed lab still weighing whether building proprietary chips is worth the investment.
Related Stories
- An OpenAI Model Broke Its Own Rules and Hacked Hugging Face in a Safety Test
- TF AI Round-up: 09 August 2026
- Nvidia Doubles AI Revenue, Stalks on China Question