Nvidia's $12.9B Hugging Face Deal: Its Biggest Acquisition Ever

Nvidia has reportedly agreed to buy Hugging Face, the open-source AI model hub, for $12.9 billion — nearly double the company's previous largest acquisition, and a huge repricing of a deal Hugging Face reportedly turned down less than a year ago.

The Information broke the story on August 26, 2026, reporting that Nvidia has agreed to buy Hugging Face, and it was quickly corroborated by CNBC, Fortune, and Forbes. As of this writing, neither company has confirmed the deal publicly, and TechCrunch reported that no signed contract is known to exist yet — meaning terms, or the deal itself, could still change before an official announcement.

Nvidia's Biggest Deal Ever, By a Wide Margin

At $12.9 billion, this would blow past Nvidia's previous largest completed acquisition — Mellanox Technologies, the Israeli networking-chip maker it bought for $6.9 billion in 2020. Nvidia has attempted bigger: its $40 billion bid for chip designer Arm, also announced in 2020, collapsed in 2022 after regulators in the UK, EU, and US signaled they'd block it on antitrust grounds. That history matters here — a Hugging Face deal of this size will draw antitrust scrutiny of its own, given Nvidia's existing dominance in AI training and inference hardware.

What makes the price striking isn't just the absolute number — it's how fast it moved. The Financial Times reported in January 2026 that Hugging Face turned down a roughly $500 million investment from Nvidia that would have valued the company at about $7 billion. Eight months later, Nvidia is reportedly paying $12.9 billion to buy the whole company outright — an increase of nearly 1.85x on a valuation Hugging Face had already rejected as too low.

What Hugging Face Actually Is

Hugging Face is the default hosting and distribution layer for open-source AI — a repository, often described as "GitHub for AI models," where developers publish, download, and fine-tune more than a million open-weight models, datasets, and benchmark suites. It monetizes on top of that free hub through Inference Endpoints (pay-per-hour GPU hosting), Enterprise Hub licenses, and Pro subscriptions, which together had pushed the company past $100 million in ARR by mid-2026 on a $4.5 billion valuation set by its 2023 Series D.

A $12.9 billion price is roughly 2.9x that 2023 Series D mark — a large jump, but one that reflects how central Hugging Face has become to the open-source AI ecosystem in the years since, not a company caught flat-footed by a lowball offer.

Why Nvidia Wants to Own the Open-Source Layer

Nvidia's core business is chips, but its moat has always extended into software — CUDA locks developers into Nvidia hardware at the compute layer. Hugging Face would extend that same logic one level up the stack, into the distribution layer where developers actually discover and deploy models. If every open-source model release funnels through a hub Nvidia owns, Nvidia gains visibility into usage patterns, influence over which inference backends get default support, and a direct channel to steer developers toward Nvidia-optimized deployment paths — reinforcing the same lock-in CUDA has produced for a decade, just one layer removed from silicon.

The timing lines up with a broader shift we've covered in how open-source AI is closing the gap with closed models. Chinese labs like DeepSeek and Qwen, along with Meta's Llama family and Mistral, have made open-weight models competitive with proprietary systems from OpenAI and Anthropic on many benchmarks. Owning the hub where that momentum concentrates gives Nvidia a hedge: if the industry tilts further toward open weights, Nvidia sits at the center of it regardless of which lab wins.

What the headline misses

This is still a reported deal, not a closed one. TechCrunch's reporting is explicit that no signed contract is known to exist, which means price, structure, or the deal itself could shift before either company confirms anything — and deals at this size routinely draw competing bids once they leak, which is itself a plausible explanation for why the price moved so far past the $7 billion Nvidia offered privately in January.

There's also a real regulatory question sitting underneath the number. Nvidia already controls the overwhelming majority of AI training and inference silicon; adding ownership of the primary open-source model distribution hub gives regulators in the US, EU, and UK — the same authorities that blocked the Arm deal in 2022 — an obvious vertical-integration argument to scrutinize. A deal announcement is not the same as a deal closing, and Nvidia has direct, recent experience with that gap.

Bottom Line

Bottom line: A reported $12.9 billion price for Hugging Face would make this Nvidia's largest acquisition ever, and it tells you how much strategic value Nvidia now places on owning the distribution layer for open-source AI, not just the silicon underneath it. The price jumped nearly 2x from what Nvidia itself offered privately eight months ago, which either reflects genuine urgency, competing interest from other buyers, or both. Until there's a signed contract and regulatory sign-off, treat the number as a strong signal of intent — not a done deal, though it would still be a modest line item next to the much larger AI megadeals defining 2026's M&A market.