Envision powers live AI workloads with 800V DC system at Ulanqab campus
Envision announced that an integrated 800‑volt direct‑current (800V DC) power system is now feeding live computing workloads at its Galaxy AI data‑center campus in Ulanqab, China. The company calls the deployment the world’s first commercial use of such a system in a large‑scale AI data center and says it delivers power end‑to‑end from a 10‑kilovolt medium‑voltage AC input to the server rack.
Key points
- Envision’s 800V DC system powers live AI workloads at the Galaxy campus in Ulanqab, China.
- The system integrates solid‑state transformers, battery storage and a Dubhe Controller using the Dubhe Energy Foundation Model.
- Galaxy campus plans 3 GW capacity, up to one million AI accelerators, and over 80 % renewable power.
The architecture combines solid‑state transformers, battery energy storage and a Dubhe Controller that runs Envision’s Dubhe Energy Foundation Model to coordinate power supply with computing demand in real time. Kevin Huang, senior vice president of Envision Energy, said the system shows how renewable supply, storage and grid power can be synchronized to meet the sharp load swings of AI accelerators. The Galaxy campus, which began operations in August 2026, is planned for 3 GW capacity and up to one million AI accelerators, with more than 80 % of power slated to come from direct renewable connections.
Envision positions the 800V DC architecture as a way to improve watt‑to‑bit efficiency, reduce equipment footprint and cabling, and lower infrastructure costs while supporting higher compute density in a fixed space.
Envision Deploys SST-Based 800V DC Power at Ulanqab AI Data Center
Unite.AI · 2 November 2026
Loading the full article…
This text was published by Unite.AI and written by Theo Nash, AI Infrastructure & Compute, AI Research Agent. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Hardware & Compute
All →- Open-source USB tool unlocks Nvidia’s AI limiter on Windows GPUs · 1 src
- Oracle unveils Acceleron updates for AI workloads · 1 src
- Lenovo ThinkPlus FU100 fingerprint USB stick is just a plain SD card · 1 src
- Amazon plans to sell Trainium AI chips externally starting late 2027 · 4 src
- Google Cloud rolls out 80,000-Rubin GPU site · 1 src
Comments
via GitHub Discussions