Nvidia's Groq Licensing Pact Draws DOJ Scrutiny as the Chipmaker Morphs Into an AI Financier

The DOJ is examining whether Nvidia's licensing arrangement with Groq was structured to avoid antitrust review, as Nvidia's AI expansion continues.

Nvidia's legal department has a new item on its desk. The U.S. Department of Justice has opened an inquiry into whether the chipmaker deliberately structured its technology licensing arrangement with AI chip startup Groq to sidestep antitrust review, according to reporting from Reuters, Bloomberg and the New York Times. A formal request for information reached Nvidia shortly after the deal was announced in December 2025.

The two accounts of the arrangement's size diverge — one puts it at $20 billion, the other at $17 billion — but the mechanics are not in dispute. Nvidia never bought Groq outright. Instead it secured non-exclusive rights to the startup's technology and brought key personnel aboard, including founder Jonathan Ross. Critics have a label for the playbook: a "reverse talent acquisition."

Unwinding the transaction is viewed as unlikely. Fines, on the other hand, remain a live possibility, and the regulatory climate stands as an uncertainty that is hard to dismiss.

A Stock That Shrugs

Investors have so far taken the news in stride. The shares closed Thursday at EUR 188.14 and were trading at EUR 189.18 on Friday, a modest gain. Over the past seven sessions the stock has shed 4.6 percent, and the monthly tally shows a decline of 2.8 percent. The 52-week high of EUR 202.50, touched in mid-May, sits 6.6 percent above the current level.

Momentum has flattened out: the price is just 2.4 percent above its 50-day moving average, with 30-day volatility running at 40 percent. Even so, the twelve-month performance remains positive at 25 percent.

Jensen Huang's Real Message

At the Goldman Sachs conference on Thursday, CEO Jensen Huang offered a line that says more about Nvidia's next chapter than any revenue forecast: the bottleneck in AI buildout is shifting from silicon to financing. He reaffirmed guidance of roughly 70 percent revenue growth for fiscal 2027 and named cybersecurity as the next major AI application area, pointing to partnerships with CrowdStrike, Cisco and Palantir. Nvidia is also putting Palantir's Foundry and Nemotron technology to work in its own supply chain — testing its AI tools internally before selling them into manufacturing, energy, healthcare and aviation.

The bolder claim is that Nvidia's systems should be understood not as depreciating hardware but as durable collateral. Morgan Stanley has a name for it — "Balance Sheet as a Service" — and has raised its price target to USD 300 while staying cautious on credit risk. The bank projects roughly USD 200 billion in lending exposure by the end of 2028, of which about USD 170 billion would be contingent liabilities from leasing, residual value guarantees and revenue-sharing arrangements.

Circular Financing or Smart Capital Allocation?

That framing invites the criticism Huang now faces regularly. The Bank for International Settlements warned in its annual report about circular AI financing as a trillion-dollar systemic risk, and investors are asking whether Nvidia's stakes in CoreWeave, OpenAI and other customers amount to buying demand that later gets presented as organic growth. Huang pushes back firmly: the investments are "immaterial" relative to the business they generate, and the company only enters "sure-win" deals. The argument holds as long as underlying demand is genuine — but the assurance alone does not shrink the balance sheet risks.

Expansion Continues Regardless

Regulatory noise has not slowed the growth machine. In early September, Nvidia signed a definitive agreement to acquire Hugging Face for USD 12.93 billion — about USD 11.9 billion in cash plus up to USD 1 billion in employee equity. The deal is expected to close in the first half of 2027, subject to regulatory approval, and Nvidia has pledged to keep the Hugging Face platform open and continue supporting other chipmakers.

The operating base for such purchases is substantial. In the second quarter of fiscal 2027, which ended in late July, revenue hit USD 96.2 billion, up 106 percent year over year, with the data center segment contributing USD 89 billion. For the third quarter, Nvidia guided to USD 108.0 billion in revenue, backed by an order backlog exceeding USD 2 trillion. The Vera Rubin platform is ramping into full production with partners including Microsoft Azure, Google Cloud and Oracle Cloud. Piper Sandler and Morgan Stanley both cite the Rubin architecture in their USD 300 targets, noting it promises thirty times the throughput per megawatt of Grace Blackwell Ultra at lower token costs.

Insider Selling and a Dividend Footnote

Alongside the strategic headlines, director Mark A. Stevens sold 1.85 million shares between late August and early September, generating proceeds of USD 410.84 million. He still holds roughly 30 million shares worth USD 6.91 billion, or 94.18 percent of his prior stake, per the required disclosures. Nvidia also declared a cash dividend of USD 0.25 per share, with an ex-date of September 10 and payment in early October.

The Quiet Problem in Gaming

Away from the spotlight, a different story is unfolding. Media reports indicate Nvidia has shipped no new gaming GPU this year — the first product pause since 1999. The planned RTX 50 Super series was halted, and the RTX 60 generation has slipped from late 2027 to 2028. A memory chip shortage is the culprit, diverting production capacity toward higher-margin AI accelerators. Given the dominance of the data center business, the financial impact is likely secondary — but strategically it underscores how thoroughly Nvidia has reoriented itself around AI.

The company is translating market power into a financial ecosystem whose complexity grows with every new deal. Operational strength is not in question. The valuation case, however, increasingly rests on whether regulators and credit markets keep endorsing the construction.