Local opposition to hyperscale AI data centers is turning into a national bottleneck: over 100 towns have imposed moratoriums and 12 states are drafting statewide bans. The backlash stems from noise, energy spikes and water use, and politicians are now demanding federal limits. This could stall AI compute expansion and pressure tech firms’ growth plans.
Chinese open‑source models such as DeepSeek’s R1, Moonshot’s K2.5 and Alibaba’s Qwen are now matching top U.S. systems while costing roughly one‑seventh as much. Their free weights have made them the most downloaded AI models globally, prompting U.S. startups to switch and threatening the U.S.‑centric innovation pipeline.
LineShine, a CPU‑only supercomputer built in Shenzhen, clocked 2.198 exaflops on the HPL benchmark, seizing the top spot on the June 2026 TOP500 list. It is the first Chinese system to lead since 2017 and marks the first sustained exascale performance without GPUs, reshaping the US‑China high‑performance computing balance.