The US Department of Justice is examining whether Nvidia's planned $12.93 billion acquisition of Hugging Face could harm competition, Reuters reported on 30 October, citing two people familiar with the matter. CNBC independently confirmed the investigation with its own source. No antitrust violation has been established.
The question at the centre of the review is not about chips. It is about distribution. Hugging Face is where AI models and training data are published and found: Nvidia says more than 18 million people and over 200,000 companies use the platform, which hosts more than three million models and 500,000 datasets. Competing chip makers, cloud providers and model developers all rely on it. If Nvidia controls that hub, investigators want to know whether it can steer discovery, prioritise its own ecosystem, or make life harder for rivals that depend on the same shelf.
Nvidia announced the deal in September and expects to close in the first half of 2027, subject to approvals. It has said the platform will stay open to competing chip, cloud and model providers.
The Hugging Face review is not isolated. According to Reuters, the DOJ is continuing an investigation begun under the Biden administration into Nvidia's acquisitions, investments and sales practices, and is looking at the $17 billion arrangement with chip startup Groq from December 2025. That deal was structured as a non-exclusive technology licence plus the hiring of a large number of Groq employees, without Nvidia buying the company — a shape regulators are now examining closely, because it delivers much of the effect of an acquisition without the filing.
Investigators are also looking at whether Nvidia offered customers larger allocations of its scarce AI chips on condition that they also buy Nvidia networking gear and Groq processors, and at the Run:ai acquisition, whose software manages GPU resources, and the purchase of SchedMD, which develops open-source software for controlling large compute clusters.
The structural concern is easy to state and hard to price. Nvidia's CUDA software stack makes switching to other hardware expensive. The company invests billions in AI firms and data centres that then buy Nvidia chips or rent compute on Nvidia hardware, so it is frequently supplier and funder at the same time. Distribution would be a third lever: not just what the models run on, but where developers find them.
The counter-argument is that Hugging Face earns money from subscriptions, cloud compute and enterprise services, and that its value to the ecosystem is precisely its neutrality — a neutrality that Nvidia has now promised to preserve in public. Regulators will want that promise turned into something enforceable, which in practice means conditions, not statements.
What to watch is timing. A second-request-style review can run past the first half of 2027 and force a restructuring, a divestiture of parts of the platform, or behavioural commitments about how models are ranked and surfaced. The Groq structure is the more interesting precedent either way: if licensing plus hiring counts as a merger for antitrust purposes, a common shortcut around the world's merger reviews closes.
Nvidia declined to comment beyond its previous statements on the deal, according to the report.
