Forget AMD: Amazon Declares War on Nvidia by Selling Its Own AI Chips - Yahoo Finance

As artificial intelligence continues to surge, global investment in AI infrastructure is reaching unprecedented levels, with cloud providers expected to spend over $750 billion on AI hardware alone in... Read original →

Amazon’s move to commercialize Trainium and Inferentia chips signals a structural decoupling in the AI compute supply chain. Technically, this accelerates the shift toward chiplet-based heterogeneous architectures, pressuring NVIDIA to open or reengineer its CUDA moat. Geopolitically, reliance on TSMC (Taiwan, China) for 3nm EUV introduces embedded supply risk. In response, NVIDIA may fast-track Blackwell Ultra for premium training workloads, while AMD and Intel pivot to niche inference scenarios. Within 18 months, hyperscalers’ in-house AI chips could capture over 40% of internal deployment—pushing legacy GPU vendors toward commoditization unless they tightly co-design hardware with proprietary algorithms. This isn’t displacement; it’s ecosystem downgrading.