- Distribution chokepoint. Hugging Face hosts roughly three million open-source AI models used by approximately 18 million developers — the dominant public repository for AI model weights — a position Justice Department staff believe could become a structural advantage if controlled by a single hardware vendor.
- Extended scrutiny. The Hugging Face probe extends a DOJ investigation into Nvidia already covering its $17 billion Groq licensing arrangement, its SchedMD acquisition and Run:ai — suggesting the department views Nvidia’s deal pattern as a systematic effort to control AI infrastructure layers.
- No action yet. A DOJ spokesperson said investigations are not evidence of wrongdoing. Nvidia expects the deal to close in the first half of 2027 and says Hugging Face will remain neutral across rival chips, clouds and software.
Reuters reported on October 28, citing two people familiar with the matter, that Justice Department staff are reviewing whether Nvidia’s $12.93 billion purchase of Hugging Face would allow the company to capture what officials described as a fundamental distribution point for AI. AI Weekly confirmed the inquiry, noting that Nvidia’s 8-K filing listed an $11.9 billion purchase price and a $1 billion employee retention pool. Nvidia declined to comment on the review; Hugging Face did not respond.
What DOJ Is Examining
The antitrust concern is structurally specific: Hugging Face is not a model builder — it is a model distribution and collaboration platform. Its library of three million open-weight models serves as the primary public mechanism by which researchers, startups and enterprises access, deploy and share AI systems. Critics of the deal argue that if a company that also competes in cloud computing, AI chips and inference hardware controlled that platform, it could steer model developers toward Nvidia-native infrastructure and away from rivals.
Nvidia has committed that Hugging Face would operate independently after the acquisition, remain compatible with competing chips and clouds, and continue to host models regardless of the hardware used to train them. A DOJ spokesperson offered the standard caveat that its inquiry does not imply any wrongdoing by Nvidia.
A Pattern Under Scrutiny
This is at least the third major Nvidia deal under active DOJ examination. An earlier probe questioned whether Nvidia structured its $17 billion Groq licensing and talent deal to avoid triggering Hart-Scott-Rodino Act review. The DOJ is also reviewing Nvidia’s December 2025 SchedMD acquisition. Taken together, the three transactions span chip supply, inference compute, workload scheduling and now model distribution — each layer of the open-source AI infrastructure stack.
The FTC sued to block Nvidia’s $40 billion Arm acquisition in 2021, and Nvidia abandoned that deal in early 2022. In September 2025, China’s market regulator separately found that Nvidia had violated its antimonopoly law, raising US-China tensions over AI hardware. The Chinese government’s own concerns about Nvidia’s market position add a geopolitical dimension to the DOJ’s domestic antitrust scrutiny.
Timeline
The Hugging Face acquisition was announced in mid-2026. Nvidia expects regulatory clearance and closing by mid-2027. The DOJ’s review is in its early stages; no subpoenas or formal second requests have been publicly reported. The inquiry was disclosed by Reuters based on sources rather than any official government statement, and the full scope of what DOJ staff are examining has not been publicly described beyond the distribution-point concern.
Jensen Huang, Nvidia’s chief executive, has separately argued in recent months that AI companies should not receive antitrust exemptions and has resisted calls for broad liability waivers for the sector — a position that puts him at odds with some AI lab executives who have lobbied for special carve-outs.