The Justice Department is investigating whether Nvidia structured its Groq deal to avoid antitrust review

The US Justice Department is examining whether Nvidia deliberately structured its licensing deal with inference chip designer Groq in a way that avoided antitrust review, according to a New York Times report.

The Justice Department opened its inquiry shortly after the deal was announced in December and has since sent Nvidia a formal request for information. If investigators find wrongdoing, they could seek fines, although forcing Nvidia to unwind the deal is considered unlikely.

The structure of the deal is what matters in this case. Nvidia took a non-exclusive licence to Groq’s chip technology and hired several of its executives, including founder Jonathan Ross, but it did not buy the company. A merger has to be reported to regulators and can be blocked. A technology licence followed by a round of hiring looks, at least on paper, like two normal business decisions.

“The Groq story is a prime example of the American system working as designed to promote innovation, reward entrepreneurs, and benefit consumers,” an Nvidia spokesperson said.

That defends the result, but it does not really answer the question at the centre of the investigation: whether the structure itself was designed to avoid scrutiny.

That may be difficult for regulators to prove as the antitrust law generally focuses on whether control of a company has changed hands, and a non-exclusive licence does not transfer control.

To make the case against Nvidia, enforcers would have to argue that the arrangement functioned as a merger, even though it was not one on paper.

The Nvidia answer is straightforward, as the licence is non-exclusive, Groq still exists, and its engineers are free to move to another company. All of that can be true while Groq is no longer in a position to compete as it did before. That distinction is likely to be central to the investigation.

What makes the case more interesting is that Groq is not the only example. Nvidia used a similar structure with Poolside, paying $6bn to license its model factory and hiring 109 staff.

Once the same pattern appears more than once, it becomes harder to treat each deal as an isolated commercial decision. For antitrust regulators, that pattern may matter more than any single transaction.

Reuters, following the NYT, puts the Groq arrangement at $17bn, and we have reported $20bn. The difference may come from whether the figure refers only to the licence payment or includes the value of the hiring arrangement, but that has not been established here. The copy should therefore avoid putting a number on the deal until the figures are reconciled.

Whatever the final figure, the effect on Groq is clearer. It had become one of the most credible independent challengers in inference silicon, after raising money at a $3.5bn valuation in a round that Nvidia itself joined.

After the deal, Groq lost its founder and senior engineers. It has since been raising $650mn to rebuild around the team and business that remain.

So while Nvidia did not acquire Groq outright, it took its technology and many of the people behind it. That leaves regulators with a question that is harder to answer than a simple merger review: when does a licence and a hiring spree amount to buying a competitor without actually buying the company?

Congress had already raised concerns about the arrangement. Senators Elizabeth Warren and Richard Blumenthal questioned the deal in March. Nvidia is also in talks with Korean inference chipmaker Rebellions, which operates in the same market but outside the reach of US merger review.

The wider significance is that Nvidia now sits at the centre of many of the AI industry’s biggest compute deals. If the Justice Department decides that this kind of licence-and-hire arrangement was used to get around antitrust scrutiny, the consequences could extend well beyond Groq.

Other companies could start using the same structure, particularly if it becomes clear that a dominant supplier can absorb a challenger without formally acquiring it.

For Europe, that may be the more important part of the story. The EU has spent the past two years developing new tools to scrutinise acquisitions of promising companies, including a referral mechanism for deals that fall below standard notification thresholds.

But a technology licence followed by a recruitment drive does not obviously fall into that category, because neither action on its own amounts to a concentration.

If the US investigation concludes that Nvidia’s structure was a way around merger rules rather than simply a different way of doing business, European regulators may have to ask whether their own rules leave a similar gap.