Nvidia has reportedly agreed to buy Hugging Face, the open-source hub where developers share and download AI models, for $12.9 billion. The Information broke the story on August 26, 2026, citing a person with knowledge of the talks, and the report has since been echoed by Forbes, CNBC, Fortune and Tom’s Hardware. Neither company has confirmed the deal publicly, and outlets including TechCrunch note that no signed contract exists yet, meaning talks could still collapse before a formal announcement.
If it closes, this would be Nvidia’s largest acquisition ever, dwarfing the $6.9 billion it paid for Mellanox in 2020 and nearly double its abandoned $40 billion bid for Arm. It would also mark a sharp turn in strategy: Nvidia, a chipmaker that built a $5 trillion market cap selling GPUs, would own the software layer where much of the world’s open-source AI development actually happens.
Nvidia’s Reported $12.9 Billion Bid for Hugging Face, Explained
According to TechCrunch’s reporting, Nvidia has agreed in principle to acquire Hugging Face for $12.9 billion, a figure that values the company at more than $13 billion once adjustments are factored in. That price is striking against Hugging Face’s own numbers: the company’s annualized revenue climbed from roughly $100 million to about $150 million in just two months this year, according to the same report. Do the math and Nvidia is paying close to 86 times current revenue, a multiple that only makes sense if you treat the deal as buying influence over the entire open-source AI stack rather than buying a software business.
The deal has not been finalized. Business Insider reported earlier in August that Hugging Face was fielding takeover interest from multiple parties, and TechCrunch’s sources caution that the agreement Nvidia reached is not yet a signed contract. That distinction matters for anyone tracking this story: as of August 28, 2026, “Nvidia buys Hugging Face” is accurate only as a reported, not a confirmed, transaction.
What Hugging Face Actually Does, and Why It Matters
Founded in 2016, Hugging Face started as a chatbot app before pivoting to become something closer to GitHub for AI models. Reuters describes it simply as “a repository of open-source AI models,” and that undersells how central the platform has become. Developers use it to publish, download, fine-tune and benchmark models, datasets and inference tools, all under one roof. When a lab releases an open-weight model, whether that is Meta’s Llama family, Mistral’s releases, or Chinese labs like DeepSeek and Alibaba’s Qwen team, it typically lands on Hugging Face first.
That neutrality is exactly what makes the platform valuable to Nvidia. Hugging Face isn’t tied to a single cloud provider or a single chip vendor. It sits in the middle of the AI supply chain as a kind of clearinghouse, and whoever owns that clearinghouse gets a say in how models get packaged, optimized and, crucially, which hardware they run best on.
Timeline: How the Deal Reportedly Came Together
The path to a $12.9 billion price tag was not a straight line. Nvidia has tried to buy into Hugging Face before and been turned down. Here’s how the numbers moved over three years.
| Date | Event | Reported Valuation | Source |
|---|---|---|---|
| August 2023 | Hugging Face closes a $235 million funding round | $4.5 billion | TechCrunch |
| Late 2025 | Nvidia offers a $500 million investment; Hugging Face declines | $7 billion (implied) | TechCrunch |
| June 2026 (approx.) | Hugging Face annualized revenue near $100 million | N/A | TechCrunch |
| August 2026 | Annualized revenue climbs to roughly $150 million | N/A | TechCrunch |
| August 26, 2026 | Nvidia reportedly agrees to acquire Hugging Face outright | $12.9 billion | The Information, via CNBC, Forbes, Fortune |
The jump from a rejected $7 billion offer in late 2025 to a reported $12.9 billion agreed price less than a year later says a lot about how fast the market for AI infrastructure “chokepoints” has moved. Hugging Face didn’t need Nvidia’s money in 2025. By mid-2026, with revenue accelerating and other suitors reportedly circling, the calculus for its board and investors clearly changed.
Why Nvidia Wants Hugging Face Now
TechCrunch’s sourcing points to three overlapping motives, and none of them is really about Hugging Face’s revenue.
- Defending chip share. OpenAI, Google, Amazon and Anthropic are each developing custom silicon to cut their dependence on Nvidia GPUs. A thriving open-source ecosystem, tuned and optimized around Nvidia’s CUDA stack, keeps smaller developers and enterprises locked into Nvidia hardware even as the biggest labs diversify.
- Re-entering cloud computing. Nvidia scaled back its own DGX Cloud rental service roughly a year ago. Hugging Face already rents out compute to developers running and fine-tuning models, giving Nvidia a ready-made path back into the cloud-rental business without rebuilding it from scratch.
- Absorbing spare capacity. Nvidia has committed billions of dollars to backstop cloud computing costs for AI customers. Idle capacity from those commitments could be redirected toward Hugging Face’s user base instead of sitting unused.
Put together, this reads less like a typical tech acquisition and more like Nvidia buying insurance against two different threats at once: custom-chip competitors chipping away at GPU demand, and cloud hyperscalers controlling too much of the distribution layer for AI models.
The Chip Independence Problem: OpenAI, Anthropic, Google and Amazon
Every major AI lab that can afford to is now building its own silicon. Google has its TPU line. Amazon has Trainium. OpenAI has been working with Broadcom on custom accelerators. Anthropic has leaned on Google’s TPUs and Amazon’s Trainium chips for large training runs. None of that spells doom for Nvidia’s GPU business overnight, but it does chip away at the assumption that every serious AI workload runs on Nvidia hardware by default.
Owning Hugging Face gives Nvidia leverage at a different layer of the stack. Instead of competing chip-for-chip with hyperscalers that can afford to design their own silicon, Nvidia would control the place where independent developers, startups and researchers, who can’t afford custom chips, actually get their models. That’s a much larger and stickier customer base than the handful of labs racing to build alternative accelerators.
Nvidia’s Retreat From, and Return To, Cloud Computing
It’s easy to miss, but Nvidia already tried running a rental cloud business once. DGX Cloud launched with partners like Oracle, Microsoft and Google to let customers rent Nvidia-optimized AI infrastructure directly. Nvidia pulled back on that push about a year before this reported deal, according to TechCrunch, choosing instead to sell hardware to cloud providers rather than compete with them directly for rental customers.
Buying Hugging Face reopens that door quietly. Hugging Face already operates paid compute services for developers who want to run inference or fine-tune models without managing their own GPU clusters. Nvidia inheriting that business avoids the awkward optics of relaunching DGX Cloud while still getting a foothold in the rental market, one layer removed from the hyperscalers it depends on for chip sales.
Regulatory Risk: What Nvidia’s Run:ai Fight With the EU Signals
Nvidia’s recent acquisition history gives regulators a clear playbook to watch for. In April 2024, Nvidia signed a share purchase agreement to acquire 100% of Israeli AI orchestration startup Run:ai. The European Commission ultimately cleared that deal unconditionally in March 2025, after reviewing it under Article 22(3) of the EU Merger Regulation even though the transaction fell below normal EU turnover thresholds. Nvidia closed the roughly $700 million purchase in December 2024.
Nvidia didn’t take that scrutiny quietly. In February 2025, Reuters reported that Nvidia sued EU antitrust regulators over their decision to accept an Italian referral scrutinizing the Run:ai deal, arguing the review process overstepped limits on regulators’ powers over smaller acquisitions. A deal the size of $12.9 billion for a platform as central to open-source AI as Hugging Face is a different order of magnitude, and would almost certainly draw a full merger review in the EU, the US, and likely the UK, focused on whether Nvidia would tilt Hugging Face’s neutral hosting toward its own hardware at the expense of AMD, Intel or custom-chip rivals.
Market Reaction and Analyst Read
Nvidia entered this news cycle as the world’s most valuable public company, with a market capitalization hovering around $5 trillion. Commentary circulating alongside the report has focused less on any single-day stock swing and more on what the price tag signals: analysts covering the deal describe the 86-times-revenue multiple as steep by any conventional standard, but defensible if Nvidia is really buying strategic control of open-source AI distribution rather than a standalone software business with $150 million in revenue.
That framing lines up with how Nvidia has approached other recent, less headline-grabbing bets on AI infrastructure. The company has previously pulled back from direct equity stakes in labs like OpenAI and Anthropic while continuing to pour money into chip supply agreements and infrastructure partnerships, a pattern that suggests Nvidia would rather own the picks-and-shovels layer of AI than bet on any single foundation model winning.
Competitive Landscape: How AI Infrastructure Deals Are Stacking Up
The reported Hugging Face price doesn’t exist in a vacuum. It follows a run of high-multiple deals for platforms that sit between AI models and the developers who use them.
| Deal | Buyer | Reported/Confirmed Price | Status (as of Aug. 28, 2026) |
|---|---|---|---|
| Hugging Face | Nvidia | $12.9 billion | Reported, unconfirmed |
| OpenRouter | Stripe | Over $7 billion | Reported |
| Run:ai | Nvidia | ~$700 million | Closed December 2024 |
| Mellanox | Nvidia | $6.9 billion | Closed 2020 |
| Arm (attempted) | Nvidia | $40 billion | Abandoned 2022 after regulatory pushback |
OpenRouter, a much smaller model-routing service, was valued at $1.3 billion in a Series B round in May 2026 before Stripe agreed to buy it for more than $7 billion just months later. That kind of rapid re-rating suggests buyers across the AI industry, not just Nvidia, are willing to pay a steep premium for any platform that sits at a chokepoint between developers and models.
Historical Context: Nvidia’s Acquisition Playbook
Nvidia’s M&A history follows a recognizable arc. The 2020 purchase of networking firm Mellanox for $6.9 billion gave Nvidia control over the interconnects that link GPUs together in data centers, a quiet but essential piece of the AI infrastructure stack. The attempted $40 billion acquisition of chip designer Arm collapsed in 2022 after regulators in the US, UK and EU objected, worried that Nvidia would use control over Arm’s licensing to disadvantage rival chipmakers.
Run:ai, at roughly a tenth the size of the reported Hugging Face deal, was Nvidia’s most recent test of how far regulators would let it reach into AI software and orchestration tools. It passed, but only after Nvidia had to fight a referral to EU merger review. A deal for Hugging Face, an order of magnitude larger and far more visible to developers worldwide, looks like the biggest test yet of whether antitrust authorities will let Nvidia extend its dominance from chips into the software layer that sits on top of them.
What This Means for Developers and Open-Source AI
For the millions of developers who currently use Hugging Face to download and share models, the immediate practical effect of an ownership change would likely be small. Model pages, download counts and community tools would probably keep functioning as they do today. The bigger question is what happens over time to hosting terms, inference pricing and, especially, whether models get quietly optimized to run best on Nvidia GPUs rather than staying hardware-neutral.
Hugging Face’s leadership has spent 2026 publicly arguing that open-weight models deserve government support rather than restriction, a position aired in television appearances discussing US policy debates over open AI models. That advocacy sits somewhat awkwardly next to a pending sale to the single largest supplier of the hardware those open models run on. Whether Hugging Face can keep functioning as a neutral hub for AMD, Intel and custom-chip rivals, or whether it drifts toward becoming Nvidia’s reference platform, is the question developers should watch most closely once (and if) a deal is finalized.
Predictions: Where This Deal Goes From Here
- Expect a formal announcement, not a collapse. Given how far talks have reportedly progressed and the public reporting from multiple outlets, a signed deal within weeks looks more likely than the talks falling apart.
- Regulatory review will stretch into 2027. Following the Run:ai precedent, expect the EU, and likely the FTC in the US, to open a formal review examining whether Nvidia would favor its own hardware inside Hugging Face’s hosting infrastructure.
- Rival chipmakers will lobby against the deal. AMD and Intel, along with custom-chip efforts at Google, Amazon and OpenAI, have a direct interest in Hugging Face staying neutral, and are likely to raise concerns during any antitrust review.
- Hugging Face’s compute-rental business will expand. Nvidia’s stated interest in re-entering cloud rental through Hugging Face suggests inference and fine-tuning pricing on the platform could shift, and new Nvidia-backed compute tiers are a plausible next step.
- More AI-infrastructure consolidation is coming. With OpenRouter, Run:ai and now Hugging Face all changing hands at steep multiples within roughly 18 months, expect other model-hosting and orchestration platforms to attract acquisition interest from cloud providers and chipmakers alike before the end of 2026.
How This Compares to Past AI Platform Acquisitions
Judged purely on dollar value, the reported Hugging Face price sits below Nvidia’s failed Arm bid but well above anything else in the company’s history. Judged on revenue multiple, it stands apart entirely. Mellanox was bought at a fraction of the multiple being reported for Hugging Face, reflecting the fact that Mellanox already had substantial, predictable hardware revenue. Hugging Face, by contrast, is being valued almost entirely on strategic position rather than current financial performance, a pattern more common in social platforms and marketplaces than in enterprise infrastructure deals.
That distinction is worth keeping in mind for anyone trying to model what Nvidia might pay for its next target. If the Hugging Face price holds, expect valuations for other developer-facing AI infrastructure platforms, from model registries to fine-tuning services, to climb in sympathy over the coming months.
Related Coverage
- OpenAI Report on Hugging Face AI Agent Hack: 4 Services Hit [2026]
- Nvidia Rubin vs AMD Helios vs Microsoft Maia 300: AI Chip Race Hits $100B [2026]
- Nvidia Earnings: $81.6B Quarter, 92% DC Surge [2026]
- Nvidia Ends OpenAI and Anthropic Investments: Inside Jensen Huang’s $40 Billion Strategic Retreat
- Cerebras CS-4 Claims 30x Faster AI Inference Than GPUs [2026]
- More AI Chips Coverage
Frequently Asked Questions
Is the Nvidia-Hugging Face acquisition confirmed?
No. As of August 28, 2026, the deal is reported but not officially confirmed by either company. The Information broke the story citing a source with knowledge of the talks, and TechCrunch reports no signed contract exists yet, meaning the agreement could still change or fall through.
How much is Nvidia reportedly paying for Hugging Face?
Reports put the price at $12.9 billion, valuing Hugging Face at over $13 billion overall, according to The Information and outlets including CNBC, Forbes and Fortune.
What is Hugging Face?
Hugging Face is an open-source platform, founded in 2016, where developers publish, download and fine-tune AI models, datasets and tools. It functions as a central repository for open-weight models from labs around the world.
Why would Nvidia want to buy Hugging Face?
Reported motives include defending Nvidia’s GPU market share against custom chips being built by OpenAI, Google, Amazon and Anthropic, re-entering the cloud compute rental business after scaling back DGX Cloud, and absorbing spare cloud capacity Nvidia has already committed to funding for AI customers, according to TechCrunch.
Did Hugging Face turn down an earlier Nvidia offer?
Yes. TechCrunch reports Nvidia proposed a $500 million investment in late 2025 that would have valued Hugging Face at $7 billion, and Hugging Face declined it.
Will regulators review the deal?
Given the size of the deal and Nvidia’s history with the Run:ai acquisition, which drew an EU merger review under Article 22(3) before being cleared unconditionally in March 2025, a Hugging Face deal of this scale would likely face review from the EU, the US and potentially the UK.
How does this compare to Nvidia’s other acquisitions?
It would be Nvidia’s largest completed acquisition, ahead of the $6.9 billion Mellanox deal in 2020, though smaller than the $40 billion Arm bid that Nvidia abandoned in 2022 after regulatory objections.
What happens to Hugging Face’s existing users if the deal closes?
No official changes have been announced. The open questions are whether Hugging Face keeps its hardware-neutral hosting model or shifts toward favoring Nvidia GPUs, and whether pricing for its compute-rental services changes under Nvidia ownership.