Green technology enterprise Envision has announced the inaugural commercial deployment of an integrated 800-volt direct current (800V DC) power system operating within a utility-scale artificial intelligence data centre campus.
The system is currently powering live computing workloads at Envision’s Galaxy AI data centre campus in Ulanqab, Inner Mongolia, delivering an end-to-end direct current power architecture from medium-voltage grid input down to high-density computing loads.
The 800V DC architecture replaces traditional multi-stage power conversion systems, improving watt-to-bit conversion efficiency while significantly reducing equipment footprint, internal cabling, and copper material requirements. These structural efficiency gains lower overall infrastructure development costs and allow data centre operators to deploy higher computing density per rack within existing spatial footprints.
Key technical, operational, and infrastructure metrics of the deployment include:
- Integrated hardware stack: Combines solid-state transformers (SST) for direct conversion from 10 kV medium voltage to an 800V DC bus, paired with battery energy storage systems (BESS) and direct-current voltage regulation modules.
- Real-time load balancing: Utilises Envision’s proprietary Dubhe Controller and Dubhe Energy Foundation Model to achieve millisecond-level coordination between transformers and battery storage, absorbing rapid GPU load swings locally.
- Campus scale and renewable mix: The Galaxy campus, which commenced initial operations in August 2026, features a planned capacity of 3 gigawatts (GW). At full build-out, the site is designed to support up to one million AI accelerators operating in parallel, with over 80 per cent of its power supplied via direct renewable energy connections.
- Global capacity roadmap: Serves as the operational baseline for Envision’s Mission Gobi initiative, which targets the deployment of 5GW of green AI data centre capacity across desert environments globally by 2030.
Kevin Huang, Senior Vice President of Envision Energy and President of Energy Storage and AIDC Power Systems, stated that establishing an operational 800V DC system demonstrates a viable end-to-end framework connecting renewable power generation directly with high-demand computing infrastructure.
Huang added that coordinating renewable generation, battery storage, and dynamic AI workloads in real time transitions power-compute integration from a theoretical concept into a commercially operational data centre standard.
