China offers tech giants cheap power to boost domestic AI chips (www.ft.com)

China is offering heavily discounted and preferential electricity deals to domestic tech giants to jump‑start local AI chip development and large‑scale model training. The move — through preferential tariffs, long‑term power contracts, and dedicated supply to industrial parks and data centers — aims to lower operating costs for energy‑intensive activities like semiconductor fabs, AI accelerator testing, and hyperscale model training. By cutting one of the largest recurring costs for both chip manufacturing and training clusters, Beijing hopes to accelerate deployment of domestic GPUs/AI accelerators, scale up training runs, and attract capital into local semiconductor and cloud ecosystems. For the AI/ML community this matters because power economics directly shapes what hardware gets built and how big models can grow: cheaper electricity reduces cost per FLOP, enabling more experimentation, larger models and continuous retraining at scale. Strategically, the policy bolsters China’s push for semiconductor self‑reliance amid export controls, risks creating global overcapacity, and raises concerns about grid stability and carbon intensity unless paired with renewable investment. Technically, expect faster iteration on custom accelerators, more on‑premise training clusters, and tighter integration between chip design, fabrication and cloud services — all of which could reshape hardware choices and competitive dynamics in AI infrastructure worldwide.