Glonce

Nvidia's $12.9bn Hugging Face deal raises antitrust concerns

businessThe Glonce editorsThe Register🔥81Glonce rating81Impact80Novelty72Relevance95Surprise72Players85Credibility82
Nvidia's $12.9bn Hugging Face deal raises antitrust concerns

Nvidia has agreed to acquire Hugging Face, the leading host for open AI models and datasets, for $12.9 billion. The Register's Dan Robinson broke the story a day earlier and sat on the podcast alongside host Brandon Vigliarolo and senior reporter Tom Claburn to discuss it. Robinson said the agreement is due to close next year provided regulators raise no objections, and that Nvidia has committed to a hands-off approach, letting Hugging Face keep operating as it does now.

Vigliarolo said Hugging Face's CEO, Clement, framed the deal as giving the company more resources to pursue its mission of reaching 100 million developers, a goal Clement reportedly worried the company could not hit without extra backing. The panel weighed that ambition against Hugging Face's current scale, estimated by the reporters at about 80 million users, against roughly 180 million for GitHub since Microsoft's acquisition of that platform, alongside GitHub's own stability problems in recent months.

Claburn questioned the economics behind the steady daily flow of new models onto Hugging Face, noting many of the companies publishing them are not generating revenue to cover the cost of building them, and suggested the acquisition could be part of the AI industry's broader pattern of circular financing. He also said a previously reported Nvidia investment in Perplexity has not actually happened and has not been publicly announced, despite earlier reports.

On potential conflicts of interest, the panel referenced an op-ed by The Register's systems editor, Tobias Mann, arguing the deal is a ripe opportunity for antitrust regulators given Nvidia's power in the AI market. Claburn and Vigliarolo were doubtful regulators would step in, pointing to the US government's recent decision not to pursue a breakup of Google over its ad-tech business, and to Meta's $17 billion social-media lawsuit settlement, a sum they called small relative to Meta's revenue and further diluted by being spread over ten years, with about 30 percent of it withheld unless YouTube and TikTok voluntarily adopt similar age-gating and youth-profile restrictions.

Looking ahead, Claburn predicted the deal will push other companies to build their own model repositories as a hedge, citing Apple's MLX framework as a likely base for Apple-optimized models and suggesting AMD could do something similar defensively. Vigliarolo compared Hugging Face's current role to "the AI Switzerland," open to any developer, and worried that fragmentation into competing platforms would be a net loss for the AI community. Claburn argued that once software and hardware are tied together this way, switching becomes harder, which is why rival hardware makers may feel compelled to respond, and that open-weight models are becoming commoditized while companies able to sustain their finances, such as Anthropic and OpenAI, focus on selling higher-end capability to government clients.

Key facts

  • Nvidia agreed to acquire Hugging Face for $12.9 billion, with the deal pending regulatory approval and expected to close next year.
  • Hugging Face's CEO, Clement, said the company wants to reach 100 million developers; Nvidia has committed to a hands-off approach to Hugging Face's operations.
  • The Register's systems editor, Tobias Mann, argued in an op-ed that the deal is ripe for antitrust scrutiny; the panel doubted regulators will act, citing the US government's recent decision not to pursue a Google ad-tech breakup and Meta's $17 billion settlement.
  • Senior reporter Tom Claburn said a previously reported Nvidia stake purchase in Perplexity has not actually happened and has not been publicly announced.
  • Panelists expect the deal to push rivals, including Apple with its MLX framework and possibly AMD, to build their own model repositories as a hedge.

Why it matters

Hugging Face is the default hosting ground for open AI models and datasets, the place new open releases and community discussion around them tend to appear first. Nvidia buying that hub for $12.9 billion hands the dominant AI hardware maker a position at the center of the open-model supply chain, not just the chips underneath it. The panel's central worry is a feedback loop: once model distribution and hardware sit under one owner, switching away from that hardware gets harder, which is exactly the kind of leverage antitrust rules exist to catch.

Who it affects

Developers who publish or download open models on Hugging Face, and the wider open-source AI community that treats the platform as neutral ground. It also affects Nvidia's rivals in AI hardware and infrastructure, since the panel expects Apple, drawing on its existing MLX framework, and potentially AMD to build their own model repositories defensively rather than depend on a competitor-owned platform. Companies such as Anthropic and OpenAI, positioned to sell higher-end models directly to government and enterprise clients, sit somewhat outside this pressure, according to the discussion.

How to use it

There is no product or pricing change here for developers today: Nvidia has said it will keep Hugging Face operating as is, and the acquisition has not closed. The practical takeaway from the discussion is a caution rather than an action: teams that depend heavily on Hugging Face for model distribution may want to watch for any tilt toward Nvidia-optimized support or documentation once the deal closes, since the panel considers that a real possibility even without any policy change being announced.

How solid is it

This is a discussion by Register staff, one of whom, Dan Robinson, broke the acquisition story and relayed what he had heard about the announcement call, giving the panel a secondhand account of how the deal was framed. Several specifics remain soft: the $12.9 billion price and the plan to close next year are stated directly, but the user counts for Hugging Face and GitHub are the reporters' own rough estimates, not cited figures, and there is no on-the-record quote from Nvidia or Hugging Face leadership beyond secondhand accounts of the call.

Risks and caveats

The deal has not closed and remains subject to regulatory approval, so terms could still change. The panel's antitrust argument is speculative: Tobias Mann's op-ed makes the case that scrutiny is warranted, but the reporters themselves doubt US regulators will intervene, given a track record they describe as weak, including the recent decision not to break up Google's ad-tech business and a Meta settlement they consider modest relative to its revenue. The claim that a separate Nvidia investment in Perplexity has not happened is Claburn's understanding as of recording and could be superseded by later news.

“Hugging Face right now is kind of like the AI Switzerland; it's a platform that allows any and all comers to host their models on there, let people pick them apart, let people try them out.”

— Brandon Vigliarolo, host of The Register's Kettle podcast