Skip to main content

Nvidia Is Buying the Home of Open Weights in a $12.9B Deal

Nvidia Is Buying the Home of Open Weights in a $12.9B Deal

August 27, 2026 · AI · Open Source · Hardware · NVIDIA

Hugging Face logo

Image: Hugging Face logo (huggingface.co).

I stopped scrolling when I saw this one. The chip company that has made billions selling compute to the AI boom is reportedly buying the place where the models that run on that compute live. Nvidia has agreed — per The Information, citing a source with knowledge of the deal — to buy Hugging Face for $12.9 billion. The home of opensource-machine-learning, where roughly 2.5 million models and 950,000 datasets live, would become part of Nvidia.

Before I get into why this matters, the responsible caveat: this is still a "reported" deal, not a signed one. Business Insider, which first reported Hugging Face fielding takeover interest over the weekend, said talks valuing the company at more than $13 billion "had not yet produced a signed agreement and could still atomize." Reuters couldn't independently verify it. So I'm writing this as a report of a report — but the logic of why Nvidia would want it is clear enough to take seriously even if the price shifts.

Why Nvidia wants the open-weights hub

Nvidia CEO Jensen Huang

Nvidia founder and CEO Jensen Huang. The open-source pivot is his bet as much as it is the company's. (photo: © European Union 2026, CC BY 4.0, via Wikimedia Commons)

The short version: protecting the moat. Nvidia's dominance in AI chips looks increasingly under pressure from exactly the companies that buy the most of them. OpenAI, Google, Amazon, and Anthropic — the biggest closed-source labs — are all building their own accelerators to lean less on Nvidia. A crowded, thriving open-source ecosystem is Nvidia's counterpunch: more open models mean more customers who will happily rent Nvidia hardware rather than feed a closed lab's in-house chip effort.

That thread has been running all year. Nvidia has poured tens of billions into funding open-source AI, and CEO Jensen Huang signed an open-weight-model letter alongside Hugging Face and 24 other companies urging the U.S. not to restrict open models. Hugging Face's own CEO, Clem Delangue, has spent months publicly pushing the same line — warning China is "clearly dominating" open-source AI and arguing the open ecosystem keeps the market from being carved up between closed labs.

Own the hub, and you own the distribution channel that decides which open models get used, fine-tuned, and deployed — and where. For a company whose revenue is compute, that is a very strategic thing to own.

The cloud comeback is the quieter part

There's a second, less sexy angle. Nvidia scaled back its own cloud business, DGX Cloud, about a year ago. But it's committed to backstopping tens of billions of dollars in cloud-compute deals for customers — and if those customers don't burn through all the compute they signed up for, Nvidia has promises to honor and idle capacity to place. Owning Hugging Face, which already lets developers rent compute to run models, gives Nvidia a natural outlet to resell that spare capacity. It's a hedge, a distribution channel, and a cloud re-entry all in one acquisition.

The price is the story

Bar chart comparing Hugging Face's $4.5B 2023 valuation to the reported $12.9B 2026 deal

Hugging Face's last known valuation to the reported deal. Nvidia was already an investor in the 2023 round. Chart by the author, press-reported figures.

MetricValue
Reported deal value$12.9B (The Information) / $13B+ (Business Insider)
2023 Series D valuation$4.5B ($235M raised, Salesforce-led)
Registered users13+ million
Models hosted~2.5 million
Datasets hosted~950,000
Annualized revenue~$150M (up from ~$100M two months prior)
Price-to-revenue implied~86× (The Information calls it pricey)

The numbers put the bet in perspective. If the ~$150M annualized revenue holds, that's roughly an 86x multiple — a rich price by any measure, and The Information itself flags it as such. It's also the latest in a spending spree: Nvidia paid ~$20 billion last December for tech and talent from inference-chip startup Groq, struck a $6 billion licensing deal with coding-startup Poolside, and bought up AI-training shops Kumo AI and Essential AI this June. Valuation-wise, this would be Nvidia paying 3x what the market cleared for Hugging Face three years ago.

What it means if it closes

For the open-source world, this is the big one. Hugging Face has positioned itself as the neutral, open-weights-forever home of machine learning — practically the default address for open models, including several I've written about this month. Making that neutral home a wholly-owned Nvidia asset changes the optics even if the model cards, downloads, and licenses stay exactly as open as they are today. There's a real question about whether the mile-wide trust that made Hugging Face the hub survives the acquisition. The counter, of course, is that Nvidia's entire strategy here depends on open models thriving — it wouldn't buy the hub to kill it.

The scale on the table

Bar chart: 13M registered users, 2.5M models, 0.95M datasets on Hugging Face

What Nvidia would be buying: the hub's early-2026 numbers, press-reported. Chart by the author.

Bar chart comparing Groq $20B, Hugging Face $12.9B reported, Poolside $6B licensing

Nvidia's recent AI deal-making, one year in. Groq was Dec 2025 tech & talent; Hugging Face is the reported deal. Chart by the author, press-reported.

My take

This is the logical endpoint of Nvidia's open-source pivot, and it's a big one. The reported $12.9 billion isn't about adding a models directory — it's Nvidia buying the distribution layer of the ecosystem it needs to keep selling chips, plus a hedge against its own idle cloud promises. For someone who cares about open weights, the thing I'd watch isn't the price, it's the trust. Hugging Face worked because it felt like a commons, not a vendor. If the acquisition closes, the open-source community gets the fastest route to Nvidia's attention and dollars — and, in exchange, loses a neutral middle. Whether that trade is worth it is the real story, and it'll play out over the next year.

Related on this blog: A Humanoid for $7,000: Physical AI Goes Open Source · SpaceX wants a million AI satellites in orbit — and they'll all run on NVIDIA

Sources: The Information ($12.9B agreed, Aug 26) via TechCrunch and PYMNTS; Business Insider ($13B+, talks not yet signed, could atomize) via TechCrunch; Reuters (could not independently verify) — all cited with the deal still unconfirmed by Nvidia or Hugging Face. Company/user/model/revenue figures are press-reported (The Information, iHeartRadio/Woodstock, PYMNTS, Fortune). Hugging Face's ~86× price-to-revenue and the ~$150M annualized revenue are as reported; I'll re-verify against any formal announcement.

Comments