Nvidia faces DOJ antitrust scrutiny over proposed Hugging Face open-source deal: What to know - The News International

DOJ's intervention into Nvidia's proposed Hugging Face deal is Washington's first systematic interception of vertical integration across AI compute and model distribution. On the technical stack, folding the world's largest open-source model repository into a CUDA-locked ecosystem would reduce chip selection from a performance race to an ecosystem lock-in, systematically compressing bargaining power across upstream EDA tooling and downstream inference frameworks. On compliance, drawing parallels to the Qualcomm-NXP aborted merger and the Intel-Movidius precedent, the DOJ will almost certainly demand divestiture of the hosting layer or impose interoperability mandates—eroding 15-20% of projected synergies in legal and operational overhead. Competitively, AMD is accelerating ROCm-Hugging Face integration while AWS Trainium teams may launch an independent model marketplace, constructing a de-Nvidia alternative pathway. Over the next 18 months, three outcomes are probable: Nvidia pivots to a licensing model to sidestep antitrust red lines; open-source AI distribution bifurcates into a US-compliance track and a sovereign-cloud track; and this case becomes the precedent anchor for all future AI-infrastructure M&A, subjecting chip-software consolidation to far stricter market-definition scrutiny.