China's AI Compute Infrastructure: GW-Scale Centers, Native Liquid Cooling, Compute-Power Synergy
智东西 · wechat · 2026-07-06
A detailed article on the new frontier of China's AI compute infrastructure. Background compares OpenAI's Stargate project challenges (funding, power, policy hurdles) with Musk's SpaceX plan for orbital compute centers. Technical section details systemic re-engineering from chips to grid: chip-side vertical power delivery reduces loop from centimeter to millimeter, supporting GPUs from 1000W to 1500W+; rack power distribution shifts to 800V HVDC, improving overall power efficiency from 80% to over 95%; cooling adopts two-phase liquid cooling to replace single-phase, supporting 3000W+ chips with PUE below 1.1; grid-side proposes compute-power synergy. Article previews OCTS2026 Open Computing Conference in Beijing on July 9.
More from Infra
- SF Compute founder: buying compute is 'an absolutely awful experience' right now — IgorCarron · 2026-09-11
- SmolVM open-sources persistent computer infrastructure for agents that outlive chat sessions — aniketmaurya · 2026-09-11
- PyTorch Day Korea 2026 launches first offline conf, CFP closes Sept 13 — PyTorch · 2026-09-11
- Local LLM server dilemma: 4x CMP-170HX (price up 53% in 20 days) vs Mac Studio M5 Ultra — rumboll · 2026-09-11
- llama.cpp lands Flash Attention tuning for RDNA4, big prefill gains on AMD — pmttyji · 2026-09-11
- Your p99 latency benchmark may be lying: a deep dive into coordinated omission — Franc0Fernand0 · 2026-09-11