US Department of Justice Opens Inquiry into Nvidia's $12.9 Billion Acquisition of Hugging Face

On October 28 local time, the U.S. Department of Justice has opened inquiries into Nvidia's $12.93 billion acquisition of the open-source AI platform Hugging Face, assessing how the deal would affect competition in the artificial intelligence sector.

Sources said the core question for Justice Department officials is whether the deal would give Nvidia control of a "fundamental distribution point" for AI models and software. Meanwhile, the department's long-running antitrust investigation into Nvidia's business practices continues, with the scope of inquiry covering its acquisitions, investments, sales tactics, and whether it has pressured customers.

Nvidia formally announced the acquisition on September 3, 2026. The consideration includes about $11.9 billion in cash to be paid to Hugging Face shareholders, plus up to $1 billion in employee retention equity incentives, with the deal expected to close in the first half of 2027.

Industry insiders know that Hugging Face is no ordinary startup—it hosts more than 3 million AI models and 500,000 datasets, serves over 18 million developers and 200,000 enterprises, and is widely regarded as the "hub" of the open model ecosystem.

A chip company that sells computing power paying nearly 90 times its annualized revenue to buy this platform, which has no strong binding relationship with computing power, is itself a signal worth unpacking.

Not Just "a Transaction"

The Justice Department's inquiry into the Hugging Face deal cannot be viewed in isolation. It is nested within a staff-led investigation that has spanned the Biden and Trump administrations and lasted for years.

Investigators have previously examined Nvidia's acquisitions of Run:ai and SchedMD, as well as the personnel and licensing arrangements with Groq, and are now checking a more damaging allegation: whether Nvidia tied priority allocation of GPUs to customers' purchases of its networking equipment or Groq processors. If true, this would mean Nvidia is using its dominance in AI chips to leverage sales in adjacent markets—a classic form of exclusive vertical lock-in.

Placing Hugging Face within this line of investigation makes the real regulatory concern clear. Hugging Face is not an ordinary asset; it is the first entry point where developers discover, compare, test, and deploy AI models. The selection habits developers form there are extremely hard to migrate once established, because alternatives would have to simultaneously replicate millions of models, hundreds of thousands of datasets, and an entire community collaboration network. Whoever controls this entry point holds structural power to influence "which models are seen, tried, and optimized."

The Justice Department wants to ask: when this entry point belongs to the world's most powerful AI infrastructure supplier, will Nvidia's competitors—whether AMD, Intel, or cloud giants developing their own chips—be systematically placed at a disadvantage in the battle for model distribution and developer attention?

From "Selling Computing Power" to "Occupying Distribution"

Nvidia's official narrative for this acquisition is protecting the open-source community. Jensen Huang said the deal will "expand global developers' and institutions' access to AI," and promised that Hugging Face will continue to support multiple chips, clouds, and AI providers. Hugging Face itself is also a capital-intensive business; petabyte-scale model storage and global distribution bandwidth require continuous investment, and Nvidia's funds can indeed remove its infrastructure-level worries. From this perspective, the acquisition has a reasonable commercial foundation.

But a reasonable foundation does not equal a complete explanation. Over the past two years, Nvidia has systematically built out open-source models—Nemotron for agents, Isaac GR00T for robotics, and Cosmos for physical AI—and these models have long been deployed on Hugging Face; the acquisition merely upgrades a partnership into an ownership relationship.

The deeper motive lies in changes in the customer structure. OpenAI's inference chip has been taped out, Anthropic has formed an internal chip team, and Google, Amazon, and Microsoft's in-house chips are all in use. The more capable major customers are of developing their own chips, the more Nvidia needs to consolidate the dependence of small and medium-sized developers and enterprises that cannot develop their own chips on its computing power—and Hugging Face is the most concentrated entry point for reaching this group of users.

Therefore, some media outlets have described the deal as a "insurance policy" Nvidia bought for $12.9 billion—insuring not Hugging Face's revenue, but Nvidia's structural position in the AI ecosystem.

The Real Situation of Neutrality

Much of Hugging Face's value is built on its platform neutrality. Developers trust it because it does not belong to any model vendor or chip vendor, and they can compare and try out offerings across vendors on the same interface.

The Register offered a sharp analogy: you would not allow an automaker to acquire the main fuel distribution channels while also buying up the system for training repair technicians, and Nvidia's acquisition of Hugging Face does almost both at once.

Nvidia promises "not to force the use of Nvidia computing resources," but this precisely exposes the limits of behavioral commitments. The real test of neutrality lies not in explicit coercion, but in default settings, search rankings, the direction of toolchain optimization, and the biases in documentation and tutorials—everyday and hidden dimensions. When the platform's owner is simultaneously the supplier of the platform's most important participants, the conflict of interest is structural and cannot be dissolved by a piece of paper.

More thought-provoking is that a year ago Hugging Face rejected a $500 million investment from Nvidia, precisely because it did not want a single investor to gain too much say. From rejecting a minority equity investment to accepting a full acquisition, this shift itself shows that in the extremely capital-intensive AI infrastructure race, an independent neutral platform trying to both "scale up" and "stay independent" is almost an impossible trinity.

The Justice Department's review may not block the deal, because a government investigation does not equal a finding of illegality, and staff-led antitrust cases often take years and may ultimately end in settlements or behavioral remedies. But the review itself has already raised a question that goes beyond the individual case: as AI's infrastructure layer—computing power, frameworks, model distribution—accelerates its concentration into a few giants in the name of "efficiency" and "ecosystem synergy," who should safeguard those "neutral layers" that carry public-interest attributes, and in what way should they be safeguarded?

If the answer is merely an open-ended promise from the acquirer, then the principle of "free use by anyone" cherished by the open-source ecosystem will increasingly depend on one company's self-restraint rather than institutional safeguards. This may be the most critical question the Hugging Face deal review leaves for AI governance.