Reuters reports DOJ inquiry into Nvidia’s proposed Hugging Face deal - NeoTeo

DOJ's probe into Nvidia's Hugging Face acquisition is not a routine antitrust filing—it is Washington drawing a hard line at the compute-to-model-distribution chokepoint. The logic mirrors the 2001 Microsoft/IE bundling case: whoever owns the infrastructure layer cannot simultaneously monopolize the application distribution layer. If the deal closes, CUDA's moat escalates from hardware lock-in to a dual lock on both silicon and the open-source model community, effectively dissolving the governance independence of Transformers and the broader PyTorch ecosystem. For AMD and Intel, this is a once-in-a-cycle strategic window. If ROCm and oneAPI complete model certification within twelve months, they can fracture developer mindshare away from the CUDA gravity well. Hyperscalers will accelerate in-house model marketplaces to decouple from single-vendor dependency. Even if the transaction is blocked, the regulatory paradigm is set: AI vertical integration is now a standing antitrust target. The open-source model community will accelerate migration toward independent foundations, and Nvidia's valuation narrative will be forced to pivot from "GPU monopoly" to "platform ecosystem"—a shift that will compress the hardware-dependency premium embedded in its multiple. The long tail of this case will reshape how capital prices the entire AI stack for the next two years.