DOJ Probes Nvidia's $20B Deal with Groq for Antitrust Issues

The United States Department of Justice has sent Nvidia a formal request for information regarding the approximately $20 billion licensing agreement established at the end of 2025 with the startup Groq, according to a September 10 report from Bloomberg. The focus of the inquiry is the structure of the deal: Nvidia paid for a non-exclusive license of Groq's chip technology and hired several executives from the rival, including founder Jonathan Ross, without subjecting the operation to the merger review process required by the Hart-Scott-Rodino Act.

Prosecutors want to understand whether the format constitutes a reverse acquihire designed to circumvent antitrust scrutiny. Senator Elizabeth Warren, who had previously called for an investigation into Nvidia in 2024, reiterated her stance that a license combined with mass hiring is a disguised acquisition. Nvidia has not commented on the information request by the time this article was published. Groq only stated that it continues to operate independently.

Changes in the M&A Landscape for AI

This case is the first direct legal test of the model that Microsoft, Google, and Amazon have employed over the past two years. Microsoft paid $650 million to Inflection AI in March 2024 for a license and absorbed almost the entire team. Google did the same with Character.AI in August of the same year, in a $2.7 billion deal. Amazon executed the same pattern with Adept in June 2024. None were treated as mergers by regulators in the United States, the United Kingdom, or the European Union at the time.

The model survived while antitrust bodies were still evaluating how to handle generative AI. That has changed. The Federal Trade Commission concluded a study in January 2025 on partnerships between hyperscalers and AI labs that already suggested a reading of de facto acquisition. The European Commission, in its competition report on AI published in May 2026, identified the licensing combined with hiring model as a practice to monitor. The UK's Competition and Markets Authority is in the midst of reviewing the Microsoft-Inflection deal under the jurisdiction of the share of supply test.

Thus, the inquiry into Nvidia does not stand alone. If the DOJ decides that the transaction required HSR notification, Nvidia may be compelled to unwind parts of the deal, relinquish licensing rights, or sell a portion of Groq's team. Such an outcome would reprice all pending acquihire combinations, including ongoing negotiations between Meta and smaller robotics labs.

Implications Beyond the U.S.

The case has implications in at least three other jurisdictions. In the United Kingdom, the CMA has already indicated that it may apply the same reasoning to Groq's licensing, given that some of the hired engineers were based in London. In the European Union, the Commission may invoke Article 22 of the Merger Regulation even with local revenue below the threshold, replicating the Illumina-Grail precedent from 2022. In Brussels, Commissioner Andreas Schwab is the most vocal proponent in this direction.

In China, the State Administration for Market Regulation monitors the situation with dual interest. If the DOJ complicates talent acquisitions by American giants, Chinese labs like Baichuan and Zhipu may gain ground in retaining engineers. Simultaneously, the SAMR is studying whether to replicate the HSR test for local operations, which would hinder consolidations within the Alibaba and ByteDance ecosystems.

A point that tends to be lost in the debate is the financial context. Nvidia generated $165 billion in revenue in fiscal year 2026 and operates with a gross margin above 75%. A $20 billion payment for Groq's license represents 12% of its annual revenue. No independent lab survives without some form of exit for talent. If the DOJ prevails, the practical effect will not be more competition; it will be less liquidity for AI founders, with the shift of innovation to less stringent jurisdictions. Funds like Andreessen Horowitz and Founders Fund already estimate that 30% of AI deals in the pipeline for 2027 may be restructured due to this inquiry.