AI data centers strain the grid: Huawei unveils source-grid-load-storage AIDC 1.0 stack
量子位 · wechat · 2026-09-28
AI data center deployment increasingly hinges on grid availability—the IEA warns 20% of planned projects could be delayed—while global DC power use is projected to nearly double to 950TWh by 2030. NVIDIA, Google and EmeraldAI formed an AI energy alliance; Huawei unveiled its source-grid-load-storage AIDC 1.0: 1280kVA Taishan UPS, multi-voltage (270V/400V/800V) Hengshan DC UPS, SmartLi 5.0 storage, TMU-based liquid cooling co-designed with Ascend servers, and prefabricated power modules cutting delivery from 7 to 3 days (SenseTime's 18MW Lingang expansion done in 45 days). Efficiency metrics are shifting too: SenseTime pushes TPW (Tokens Per Watt), claiming 80% more tokens per unit of power cost.
More from Infra
- Sarvam AI Launches Token Factory: India-Hosted Inference for Open Models — itsOmSarraf_ · 2026-11-16
- BIS Issues Interim Final Rule Addressing Remote Access to Advanced AI Chips — apples_jimmy · 2026-11-16
- Jailbroken PS4 runs a 9B open-source model at up to 43 tok/s by talking directly to its 2013-era GPU — AIFlow_ML · 2026-11-16
- Mobile AI Factory: agents auto-build on-device SLMs with 19x less GPU time — pmttyji · 2026-11-16
- Stratechery's Fictional Obituary: Imagining Jensen Huang's 2026 Exit From Nvidia — Gregory_C_Allen · 2026-11-16
- vLLM serves DeepSeek V4.1-Flash at 954K tokens/s per GPU, $38.96/GPU-hour on GB300 NVL72 — bookwormengr · 2026-11-16