DOJ Probes Nvidia's Twenty Billion Dollar Groq Deal Over Antitrust Structure

The Justice Department has opened an investigation into whether Nvidia structured its roughly twenty billion dollar Groq transaction to avoid federal antitrust review, putting intense scrutiny on one of the largest AI deals in industry history. The probe, first reported by The New York Times on September 9 and confirmed by Bloomberg and Reuters the following day, focuses on whether the deal's unusual structure was designed to sidestep mandatory merger notification requirements.

Nvidia announced on December 24, 2025, that it would pay approximately twenty billion dollars in cash to acquire assets from Groq and secure a non-exclusive license to Groq's inference technology. The transaction saw founder and CEO Jonathan Ross, president Sunny Madra, and other senior leaders move to Nvidia, while Groq itself continued operating as an independent company under CFO Simon Edwards as the new chief executive. Reuters reports the license component alone is valued at seventeen billion dollars, making this the largest transaction Nvidia has ever completed.

The investigation hinges on a critical piece of deal structure: no company was acquired, no voting securities changed hands, and the two things that transferred—a license and a group of employees—are generally not the type of assets Hart-Scott-Rodino merger review rules were designed to capture. The probe has sent Nvidia a formal demand for information and could result in financial penalties if regulators conclude the transaction was improperly structured, though unwinding the deal is not expected.

The Non-Exclusive License Loophole

Under federal premerger notification rules, an exclusive patent license is treated as a transfer of assets and potentially reportable to antitrust authorities. A non-exclusive license, however, is not regarded as an asset for Hart-Scott-Rodino purposes at all. This technical distinction becomes the load-bearing word in Nvidia's defense: by structuring the Groq arrangement as a non-exclusive license rather than an outright acquisition, the company may have avoided triggering mandatory antitrust review entirely.

The Justice Department's question is whether what Nvidia bought was, in substance, an acquisition that had to be reported and cleared before closing, or a license-plus-hiring arrangement that never triggered the notification obligation. The 2026 size-of-transaction threshold stands at 133.9 million dollars, meaning the twenty billion dollar figure was never in doubt—what matters is whether the instrument itself qualified as a reportable transaction.

This is not the first time federal regulators have challenged creative deal structuring. On July 13, 2026, the FTC extracted penalties from Edwards Lifesciences and Genesis MedTech over a transaction split into two pieces that individually sat below reporting thresholds. That precedent suggests regulators are watching closely for arrangements designed to avoid scrutiny through technical maneuvering.

Hugging Face: The Acquisition That Followed the Rules

In stark contrast to the Groq arrangement, Nvidia's acquisition of Hugging Face—announced September 2, 2026—was structured as a traditional merger agreement complete with regulatory approval conditions. Nvidia agreed to pay 11.9 billion dollars to Hugging Face stockholders plus up to one billion dollars in equity-based retention for employees who join the company, bringing the total deal value to 12.93 billion dollars.

The Hugging Face merger agreement explicitly includes a condition requiring "receipt of required regulatory approvals" and is expected to close in the first half of 2027. This structure represents exactly what the Groq deal lacked: formal notification to antitrust authorities, a waiting period for review, and a regulatory clearance condition before the transaction can complete.

The difference is telling. When asked why Hugging Face required traditional merger clearance while Groq did not, the answer lies in what was being purchased. Hugging Face is a whole-company acquisition with voting securities changing hands—unambiguously an asset purchase under merger review rules. The Groq arrangement, structured around a non-exclusive license and employee transitions, occupies murkier territory.

Open-Source Strategy at Stake

Nvidia's AI licensing spree extends beyond Groq. The company has committed tens of billions to securing access to open-source foundation models and AI platforms, viewing them as essential drivers of demand for its GPU products. In the company's most recent risk factor disclosures, Nvidia explicitly states that "demand for open-source foundation models and applications based on them promotes the use of our products worldwide."

The threat Nvidia identifies is government restriction on open-weight models, including those originating in China. By acquiring or licensing access to the leading open-source AI platforms, Nvidia positions itself at the center of what it views as the future of AI development—a future that depends on researchers, developers, and enterprises building on freely accessible models rather than proprietary ones.

The investigation comes at a pivotal moment for AI consolidation. If the Justice Department concludes Nvidia structured the Groq deal improperly, it could establish new precedent for how AI licensing and talent acquisitions are regulated. The outcome may determine whether other tech giants can follow similar paths or whether such arrangements will face heightened scrutiny going forward.