Case Studies
Real-world 800V DC and HVDC deployments in AI data centers and grids
Envision Ulanqab Xinghe Base – First Commercial 800V SST
3 GW planned; first-phase 800V SST + BESS + DC/DC end-to-end integration (2026-11-02); 10kV → 800V; green power >80%
Foxconn Kaohsiung K-1 – NVIDIA 40MW 800V Demonstration Site
40 MW demonstration site announced 2025-10, operational 2026-05 on 800 VDC
NVIDIA Kyber / Vera Rubin – 800V Native Rack Architecture
576 GPUs per rack; 450kW per rack (Rubin generation); 800V HVDC native
Microsoft Mt Diablo & Grid-to-Chip
Medium-voltage SST from 13.8-34.5kV to 800V / ±400V DC; grid-to-chip power architecture
Microsoft First Production GB300 NVL72 Cluster
First production-grade GB300 NVL72 cluster with more than 4,600 racks
Google Project Tembo / TPU v7 – ±400V Route
Whole-rack ±400V DC approach for TPU v7 generation
xAI Colossus – Memphis AI Data Center
150MW cluster built in 122 days; gas turbines; 480V AC baseline before HVDC transition
China Telecom Qingyang – East-Data-West-Computing 800V HVDC
800V HVDC deployment with PUE 1.192 (reported 2026-05)
Hohhot Ascend 950DT 10k-GPU Pool
50.88MW AI compute facility with B12 liquid cooling
SGCC Zhejiang Jinshuitan Green AI Computing Center
104.3P computing capacity; PUE 1.13; computing-power-grid coordination
Figures are as reported by the cited sources at the given dates; verify current status before procurement decisions.