After China, OpenAI Chips Away at Nvidia: So Why is NVDA Stock Up? - Yahoo Finance

Following China's successful development of a large AI model without Nvidia chips, OpenAI has further reduced its reliance on Nvidia’s chips by cutting inference costs through software optimization an... Read original →

OpenAI’s move to reduce reliance on NVIDIA chips appears disruptive but actually reinforces a bifurcated AI hardware stack: CUDA remains unchallenged in training, while inference fragments rapidly. This technical cascade pushes TSMC to prioritize sub-3nm EUV capacity, raising barriers for smaller players. Geopolitically, U.S. export controls on advanced semiconductor tools inadvertently strengthen NVIDIA’s pricing power in compliant markets, though its supply chain concentration in Taiwan, China poses latent risk. In response to Google and Amazon’s ASIC surge, NVIDIA is fortifying its moat with Rubin architecture and NVLink interconnects—difficult to displace short-term. Over the next 12–24 months, even as custom chip shipments overtake GPUs, NVIDIA will retain premium positioning in general-purpose AI training. The real threat isn’t hardware substitution—it’s the tipping point where developer loyalty to CUDA begins to erode.