The U.S. Justice Department is investigating whether Nvidia's December licensing transaction with AI-chip startup Groq was structured to avoid antitrust scrutiny, according to a New York Times report cited by multiple outlets. Nvidia has already received a formal request for information.
A license that walked, talked and moved like an acquisition
Groq presented the December deal as a nonexclusive license giving Nvidia access to chips optimised for AI inference. As part of the same arrangement, Groq founder and chief executive Jonathan Ross and chief operating officer Sunny Madra moved to Nvidia, while Groq itself remained a separate company. Later reports put the combined package in the $17-20 billion range. On paper it is a license; in practice it looks a lot like the executive-poaching-plus-technology-in transactions that have quietly become the AI industry's alternative to a notified merger.
Why regulators care now
Nvidia sits in the multi-trillion-dollar range, and its CUDA software remains the default for training large models. A license-and-talent structure that dodges the Hart-Scott-Rodino filing is exactly the pattern lawmakers have started calling a stealth takeover. Even if the DOJ ultimately imposes no remedy, the information demand forces Nvidia to document how it folds inference rivals into its stack. Nvidia said the Groq story is "a prime example of the American system working as designed."
Startups should reprice their exits
For every inference-ASIC startup — and for well-funded challengers like Positron, which just raised $875M — the probe is a warning about exit math. Selling technology and executives to the incumbent may no longer be a quiet alternative to a notified merger. For hyperscalers looking to diversify away from Nvidia, the case may slow future absorptions but does not by itself create a second training ecosystem.
Reporting based on coverage from The New York Times and Reuters.