- Lead. The Justice Department is investigating whether Nvidia structured its $17 billion licensing deal with AI chip startup Groq to avoid premerger notification requirements under the Hart–Scott–Rodino Act, according to reporting by Bloomberg and the New York Times.
- Fact. Struck in December 2025, the arrangement saw Nvidia license Groq’s inference chip technology and hire several of its senior leaders — including founder Jonathan Ross — without technically acquiring the company.
- Stake. The probe is the first formal legal test of what some analysts are calling the “reverse-acquihire” playbook, under which large technology companies gain control over key assets and talent through licensing rather than mergers, potentially bypassing antitrust scrutiny.
Bloomberg and the New York Times first reported in September that the Justice Department had sent Nvidia a formal request for information, opening an inquiry into whether the licensing arrangement with Groq was structured to circumvent the Hart–Scott–Rodino Act’s premerger notification requirement. The HSR Act generally requires companies above certain asset thresholds to notify regulators before completing transactions that transfer control, allowing authorities time to assess competitive effects. The question for investigators is whether Nvidia obtained enough control over Groq’s technology and key personnel to constitute an effective acquisition while avoiding the formal review process, according to the Bloomberg report.
The deal’s structure
Nvidia’s arrangement with Groq was announced in late 2025. Nvidia took a non-exclusive license to Groq’s inference chip technology and brought several of Groq’s senior leaders to Nvidia, including founder Jonathan Ross. Groq itself continues to operate as a separate company. Antitrust law generally focuses on whether control of a company has changed hands; a non-exclusive licence does not transfer control in the traditional sense, and that structural choice is at the center of the DOJ’s investigation. Nvidia called the deal “a prime example of the American system working as designed to promote innovation, reward entrepreneurs, and benefit consumers.” The DOJ declined to comment on the pending matter.
Congressional scrutiny and legal risk
Senators Elizabeth Warren and Richard Blumenthal sent Nvidia a letter in March 2026 asking whether the deal’s structure was intended to avoid antitrust scrutiny, signaling early political attention to the transaction. Legal analysts say the probe could end without action — prosecutors would need to show that the arrangement’s structure, not its business purpose, determined how it was organized. Forcing Nvidia to unwind the deal is widely considered unlikely even if the DOJ finds cause to act; fines would be a more probable outcome. The investigation adds to a broader regulatory moment for AI companies: a separate class action, Buist et al. v. Anthropic, filed September 18, 2026 in the Northern District of California, alleges that Anthropic, OpenAI, Google and SpaceXAI agreed to slow AI development, citing reported coordination meetings since July 2026. That case remains at the complaint stage.
Implications for AI M&A
The DOJ’s inquiry arrives as large technology companies have accelerated their investment in AI infrastructure and talent through structures that do not require formal merger review. The Nvidia–Groq probe, alongside the class action against AI labs alleging a coordination agreement among leading AI companies, suggests regulators and plaintiff attorneys are testing the boundaries of antitrust doctrine in an industry that has grown faster than the legal frameworks designed to govern it. How the DOJ resolves the Nvidia investigation will set a precedent for whether licensing structures can substitute for acquisitions in high-stakes AI transactions.