Why Chinese AI Models Should Worry Nvidia, Micron Stock Investors - Trefis

As Chinese AI models continue to gain performance and efficiency, the global AI infrastructure investment landscape is undergoing a profound transformation. Companies like NVIDIA and Micron have long ... Read original →

Chinese AI models’ leap in inference efficiency is triggering a cascade across the semiconductor stack. Architectures like MoE and MLA in GLM-5.2 drastically cut reliance on HBM and premium GPUs, eroding NVIDIA’s and Micron’s pricing power in AI infrastructure. This forces upstream chip design toward granular energy-efficiency metrics and pushes cloud providers to deploy low-cost inference clusters. Geopolitically, U.S. export controls on EUV tools and HBM have inadvertently accelerated Chinese adoption of model distillation and algorithmic compression—reducing supply chain exposure. In response, NVIDIA may fast-track FP8-optimized successors to Blackwell, while Micron could slash HBM prices to defend share. Over the next 12–24 months, sustained Chinese model adoption in enterprise inference will shift global AI capex from brute-force compute toward algorithmic elegance, undermining valuations built on 3nm GPUs and HBM demand.