Nvidia Hugging Face Deal Faces DOJ Questions on Competition

Key Takeaways

• Nvidia agreed on 3 September 2026 to buy Hugging Face for $12.93 billion, with the deal expected to close in the first half of 2027.
• CEO Jensen Huang pledged that the platform would stay open to third-party processors, frameworks, and cloud services.
• Reuters reported on 28 October 2026, citing two sources, that the US Department of Justice is asking whether the deal lets Nvidia capture a key distribution point for AI models, leverage commercial negotiations, or disadvantage rival chipmakers.
• The questions sit inside a broader DOJ probe of Nvidia, covering Run:ai, SchedMD, the Groq arrangement, and sales practices, running since the Biden administration.
• An inquiry is NOT evidence of wrongdoing and may not lead to legal action. The deal is proposed, NOT closed.
• Both the Department of Justice and Nvidia declined to comment on the inquiries, while Hugging Face did not respond to requests for comment, according to the reporting.

A proposed deal, an inquiry, and a distinction that matters

The most important distinction in this story is also the easiest to lose in a headline. The US Department of Justice is asking questions about Nvidia’s proposed $12.93 billion acquisition of Hugging Face. Asking questions is not the same as filing a case, finding wrongdoing, or blocking a deal. An antitrust inquiry does not constitute evidence of wrongdoing and may not culminate in legal action, as Reuters itself noted in the reporting summarised here. And the deal itself is proposed, not closed. It was agreed on 3 September 2026 and is expected to close in the first half of 2027, subject to the reviews that any transaction of this size must pass. Keeping those two facts in view, inquiry versus verdict, and proposed versus closed, is essential to understanding what is actually happening.

That caution does not make the story unimportant. Hugging Face is not a minor asset. It operates a central repository where developers host, share, and test open-source AI models, code, and datasets. One briefing source describes it as the largest platform for uploading and downloading AI models, hosting 3 million models and used by 18 million developers. Whether those figures are exact is a matter for Hugging Face’s own reporting, but the scale is not in doubt: for a large part of the open AI ecosystem, Hugging Face is where models are found, tried, and shared. Nvidia, meanwhile, is the dominant supplier of the chips on which much of modern AI is trained and run. Putting the dominant chip supplier together with the central model hub is, on its face, a combination that any antitrust authority would want to examine. The question is not whether the DOJ should ask questions. It is what the answers will show.

What the DOJ is asking, according to Reuters

Reuters reported on 28 October 2026, citing two people familiar with the matter, that the Department of Justice is questioning whether Nvidia’s planned acquisition could harm competition in artificial intelligence. The Silicon Report summary of the Reuters reporting lists three specific lines of inquiry. First, whether the acquisition would let Nvidia capture a fundamental distribution point for AI. Second, whether Nvidia could use the platform as leverage in commercial negotiations. Third, whether it could harm competing chipmakers in the AI market, or disadvantage alternative model repositories. CNBC reported that it independently confirmed the inquiry with an unnamed source, which adds a second outlet’s confirmation that an inquiry exists, even though the underlying questions remain, for now, reported rather than officially published.

Each of those three questions deserves unpacking, because they point to different theories of harm. The distribution point question is about control of the gateway. If most developers discover and download models through Hugging Face, then owning Hugging Face could, in theory, give Nvidia influence over which models are prominent, which are easy to deploy, and which hardware they are optimised for. The leverage question is about bargaining power. If a customer or partner depends on Hugging Face for model distribution, Nvidia could, in theory, use that dependence to gain better terms in chip sales or other commercial talks. The rival chipmaker question is about foreclosure. If models on Hugging Face work best, or only, on Nvidia hardware, rival chipmakers could find it harder to compete, even if their hardware is technically competitive. None of these theories has been proven. They are questions that investigators are asking, and the answers will depend on facts about how Hugging Face actually operates, what Nvidia actually plans, and what safeguards, if any, are put in place.

It is also important to note what the reporting does not say. It does not say that the DOJ has decided the deal is anticompetitive. It does not say that a lawsuit is imminent. It does not say that Nvidia has done anything improper in its handling of Hugging Face so far, because Nvidia does not yet own Hugging Face. The deal has not closed. The inquiry is, at this stage, an examination of a proposed future, not a judgement on a completed past.

What Nvidia has pledged, and why pledges are tested

When Nvidia announced the agreement on 3 September 2026, chief executive Jensen Huang stated that the platform would remain open to third-party processors, frameworks, and cloud services. That pledge is central to the deal’s public case. Nvidia’s argument, in effect, is that Hugging Face’s value lies in its openness, and that closing it off to rivals would destroy the very asset Nvidia is buying. If developers left Hugging Face because it became a Nvidia-only shop, the platform would lose its network effect, and Nvidia would be left with an expensive, diminished hub.

Antitrust investigators, however, do not simply accept pledges at face value. They test them against incentives, history, and enforceability. A pledge to remain open can be honoured in letter but narrowed in practice, for example by optimising new features for one hardware family first, by making integration with rival hardware slower or less well documented, or by bundling Hugging Face access with chip purchases in ways that are difficult to refuse. None of those behaviours is alleged in the current reporting, and this article does not allege them. The point is narrower: investigators will want to know not just what Nvidia says it will do, but what mechanisms would prevent it from doing otherwise, and what would happen if it did. That is standard antitrust practice, and it is why a pledge, however sincere, does not end an inquiry.

The broader context also matters. Hugging Face is used by developers who work across many hardware platforms, many cloud providers, and many model families. For those developers, the question is practical, not theoretical. Will their models still run well on non-Nvidia hardware? Will their deployments still work on the cloud of their choice? Will the platform’s rules change in ways that favour one ecosystem? Those are the questions that will determine whether the open AI community continues to treat Hugging Face as neutral ground, or begins to look for alternatives. The DOJ’s inquiry, whatever its legal outcome, has already brought those questions into the open, and developers are entitled to ask them of Nvidia directly.

The broader DOJ probe of Nvidia

The Hugging Face questions do not sit in isolation. They form part of an expansive Justice Department investigation into Nvidia’s business conduct, according to the Silicon Report summary. That review originated during the Biden administration and has continued under the Trump administration, with Reuters reporting no indication of White House intervention. The broader inquiry reviews Nvidia’s acquisitions, investments, sales practices, and whether the company pressured its customers. Investigators have also examined whether Nvidia allocated more AI chips to customers that purchased its networking gear alongside processors, as well as its acquisitions of Run:ai and SchedMD, which closed in December 2025.

Enforcers are separately reviewing whether Nvidia structured an arrangement with chip startup Groq to sidestep regulatory merger reviews. Under that deal, Nvidia paid to license Groq’s technology and hired its key engineers and executives rather than purchasing the company outright, according to reporting by Bloomberg cited in coverage of the probe. Nvidia has also held discussions with Reflection AI, an open-model startup it already backs, regarding a potential acquisition or an expanded investment stake. Beyond federal scrutiny in the United States, Nvidia faces an antitrust investigation from regulators in France.

That breadth matters for two reasons. First, it means the Hugging Face inquiry will be read, by investigators and by the market, in the context of a pattern, not as a single transaction in a vacuum. A company with no prior scrutiny might face a narrower set of questions. A company already under a broad probe will face questions about how each new deal fits into the larger picture. Second, it means the timeline is uncertain. A broad probe that has run across two administrations is not a quick process, and the Hugging Face deal’s expected closing in the first half of 2027 may be affected by how long the various reviews take. One briefing source notes that if antitrust scrutiny drags on, the closing could easily slip beyond 2027. That is a possibility, not a prediction, but it is a realistic one for a deal of this size and sensitivity.

It is equally important not to overread the broader probe. The fact that investigators are examining sales practices, chip allocation, and past acquisitions does not mean they have found violations in any of those areas. Investigations examine; they do not, by themselves, prove. Both the Department of Justice and Nvidia declined to comment on the inquiries, while Hugging Face did not respond to requests for comment, according to the reporting. That silence is normal during an ongoing inquiry, and it should not be treated as an admission by any party.

Why Hugging Face matters to the open AI ecosystem

To understand why this deal attracts so much attention, it helps to understand what Hugging Face actually does. At its simplest, it is a place where developers upload models, download models, test them, and share the code and datasets around them. That sounds mundane, but in practice it is infrastructure. Just as a package registry is infrastructure for software developers, a model hub is infrastructure for AI developers. The hub reduces friction: instead of each team hosting its own models, negotiating its own downloads, and building its own testing tools, it can use a shared platform that everyone already knows.

That shared character is what makes the ownership question sensitive. If a hub is owned by a neutral party, developers can reasonably assume that it will treat all hardware and all models fairly. If it is owned by a dominant hardware supplier, developers may worry, fairly or not, that fairness will erode over time. Even if Nvidia honours its openness pledge perfectly, the perception of a conflict of interest could push some developers to diversify, to mirror their models elsewhere, or to build alternative hubs. Perception, in platform economics, can become reality, because platforms depend on trust as much as on technology.

Our own coverage of Hugging Face illustrates its range. Hugging Face Carbon-X: The Free AI Model That Lets You Chat With DNA shows how the platform hosts specialised, open models far beyond the mainstream chatbots that dominate headlines. NVIDIA Nemotron Transcribe: Open Speech Model, 40 Languages shows Nvidia’s own open-model work, and how it already intersects with the open ecosystem. And Nvidia and AMD Invest $4 Billion in Meta as It Raises $95 Billion for AI shows how Nvidia’s capital, like its chips, is woven through the AI industry in ways that go beyond a single transaction. A fourth relevant piece, FuriosaAI $620M Series D: Inference Chips at $2.9B Value, is a reminder that rival chipmakers exist, are raising significant money, and have a direct interest in whether the model hub they depend on remains neutral.

What happens next, and what would change the story

The next steps are procedural, and they will be slower than the headlines. Regulators in multiple jurisdictions must review the deal, because Hugging Face operates globally and Nvidia sells globally. The US DOJ inquiry is one part of that process, but not the only part. The deal cannot close until those reviews are complete, or until any conditions imposed by regulators are satisfied. If regulators impose conditions, they could take many forms: commitments to keep the platform open, to maintain support for rival hardware, to keep pricing and access terms non-discriminatory, or to report regularly on compliance. If regulators block the deal, which has not been suggested in the current reporting, Nvidia and Hugging Face would have to decide whether to challenge that decision, renegotiate, or walk away. None of those outcomes is currently predicted by the reporting, and this article does not predict any of them.

What would change the story is evidence, in either direction. Evidence that Nvidia plans to keep Hugging Face fully open, with enforceable commitments and independent oversight, would weaken the theories of harm. Evidence that the platform’s openness is already narrowing, or that commercial terms are being tied to chip purchases, would strengthen them. Official statements from the DOJ, Nvidia, or Hugging Face would also change the story, by moving it from reported questions to stated positions. Until then, the story remains what it is: a major proposed acquisition, under serious but preliminary scrutiny, with a pledge of openness that investigators are testing.

For developers and businesses that use Hugging Face, the practical advice is not to panic, and not to ignore the story either. Continue using the platform as you do today, because nothing has changed in its ownership or its terms yet. At the same time, keep an eye on the review process, on any conditions that are imposed, and on whether the platform’s support for your hardware and cloud choices remains as strong as it is today. Diversifying where you host critical models, keeping your own copies, and avoiding dependence on a single hub for irreplaceable assets are prudent steps in any case, regardless of who owns the hub.

The bottom line

Nvidia’s proposed $12.93 billion acquisition of Hugging Face, agreed on 3 September 2026 and expected to close in the first half of 2027, is now the subject of DOJ questions about whether it could let Nvidia capture a key distribution point for AI models, leverage commercial negotiations, or disadvantage rival chipmakers. Those questions, reported by Reuters on 28 October 2026 and confirmed in part by CNBC, sit inside a broader DOJ probe of Nvidia that covers Run:ai, SchedMD, the Groq arrangement, and sales practices, and that has run since the Biden administration. Jensen Huang has pledged that Hugging Face will stay open to third-party processors, frameworks, and clouds, and that pledge will now be tested by investigators and by the market. An inquiry is not evidence of wrongdoing, and it may not lead to legal action. The deal is proposed, not closed. What happens next will depend on facts that have not yet been made public, and on reviews that have not yet been completed. Until they are, the honest summary is that a very important platform may change hands, that serious questions are being asked about that change, and that the answers are still to come.

Sources: Silicon Report, DOJ asks how Nvidia’s $12.9B Hugging Face deal affects AI competition; LetsDataScience, Nvidia Hugging Face $12.93 billion DOJ asking questions.