DOJ probes whether Nvidia's Groq deal dodged antitrust review
Two regulatory threads and one infrastructure bet: Washington is asking whether Nvidia's biggest chip deal was structured to stay under the merger radar, Massachusetts just made data centers pay for their own clean power, and JD.com is building a 100,000-GPU cluster on Chinese silicon.
The Justice Department is investigating whether Nvidia structured its licensing deal with AI chip startup Groq to avoid antitrust scrutiny, the New York Times reported, citing two people familiar with the matter. Nvidia announced the $17 billion arrangement last December as a "non-exclusive license" to Groq's chip technology and hired several of its executives, including founder Jonathan Ross — a shape that let both companies describe it as something other than an acquisition. Per the report, the DOJ opened its inquiry shortly after the deal was announced and has since sent Nvidia a formal request for information; the agency could fine the company if it finds the deal was mishandled, though it is unlikely to try to unwind it.
The structure is the whole question. A licensing agreement plus executive hires delivers most of what an acquisition delivers — the technology, the talent, the competitive threat removed — while sidestepping the review an acquisition triggers. An Nvidia spokesperson framed it as "a prime example of the American system working as designed to promote innovation, reward entrepreneurs, and benefit consumers," and Groq and the DOJ did not respond to requests for comment. What makes this worth watching is not the size of the deal but the precedent: if talent-plus-license deals become the standard way to absorb a rival, the merger review regime gradually stops covering the transactions it was written for.
Massachusetts is now the third state in as many months to clamp down on data center development, with an executive order from Gov. Maura Healey requiring facilities above 25 megawatts of peak demand to bring their own power and meet the state's clean-energy requirements. Developers who don't generate that power onsite will need to fund new generation nearby or pay into a ratepayer protection fund. Healey also paused applications for a data center sales tax exemption that only took effect last month, and the order directs communities to avoid signing non-disclosure agreements with developers — a direct shot at the secrecy that has made local opposition so explosive.
The 25MW threshold is the detail that matters: it exempts ordinary enterprise facilities and lands squarely on AI training and inference campuses, which is to say the order is aimed at exactly the buildout driving the current boom. Pennsylvania's version of this fight arrived by ballot; this one arrives by executive order, which is faster to impose and faster to reverse. We covered the political version of this in September — Pennsylvania voters move AI data centers to the top of the ballot.
JD.com used its JDDiscovery conference in Beijing to commit to a 100,000-GPU cluster built on Chinese silicon, a joint project with chipmaker Moore Threads that the company says is the first deployment of domestic GPUs at that scale inside a major Chinese AI cloud. Alongside it came JoyAI-Echo WM, a new world model that pairs native audio-video generation with user control, and a physical-AI push that includes 10 million hours of embodied video data collection over two years, more than 80 robotics industry hubs over five years, and 10 billion yuan in retail resources committed by 2028.
The logistics arm offered the most concrete number: its Meta Brain 3.0 model cut route optimization across hundreds of millions of parcels from minutes to seconds, which is the kind of claim JD can actually verify against its own delivery network. The strategy is coherent — a retailer with warehouses, delivery routes, and a supply chain has real-world data that pure model labs don't — but the six "world's largest" targets announced on stage are corporate ambition, not results. The cluster is the part that can be checked.
What to watch: whether the DOJ's Nvidia inquiry produces a formal complaint or quietly closes, as licensing-structured deals have largely done before.
Should talent-plus-license deals face the same merger review as acquisitions? Tell us in the comments.
