DOJ Probes Nvidia’s Hugging Face Deal Over AI Competition Concerns - PYMNTS.com

The DOJ probe is less about one transaction and more about establishing a precedent: vertical integration across the AI stack—silicon to model distribution—will now face Sherman Act scrutiny. Hugging Face controls the on-ramp for the vast majority of open-source model deployment; Nvidia commands the dominant share of training compute. Merging these layers replicates the x86 lock-in playbook Intel and Microsoft executed in the 2000s, systematically narrowing viable paths for AMD, Huawei Ascend, and Google's TPU ecosystem. If structural remedies are imposed, CUDA's software moat and Hugging Face's model registry would be forcibly decoupled, inflating R&D coordination costs by an estimated 30% or more. For competitors, this is a strategic gift: Microsoft will accelerate Azure AI Foundry's independence, while AMD can package ROCm with unencumbered open-source models as a no-strings alternative. The 18-month outlook is unambiguous: the compute-model-application trilect will become the default regulatory unit globally. Any cross-layer M&A in AI infrastructure will face DMA-style ex-ante review, fundamentally rewriting how companies structure their AI roadmaps. The era of buying your way to the top of the stack is over.