Envision says 800V power system is running at Galaxy AI data center
The company describes equipment supplying live computing workloads in China. Its much larger campus-capacity figures remain plans for full build-out.
An AI campus announcement can contain two very different stories: equipment operating today and capacity a company hopes to build. Envision's latest power-system announcement is a reason to keep those columns separate.
In a November 2 release, Envision said an integrated 800-volt direct-current system is supplying live computing workloads at its Galaxy campus in Ulanqab, China. The announcement describes a connection from grid input through to computing loads. These are company-reported deployment claims, rather than independently verified operating results.
Conversion and buffering do different jobs
Envision says the system combines solid-state transformers, battery storage and controls. The transformers convert incoming power, while the storage and controls help manage changing computing loads.
That distinction matters when reading an infrastructure announcement. Delivering power in the form equipment needs and responding to changes in demand are related tasks, but evidence about one does not automatically settle the other. A deployment reference is useful; measured performance is the next question.
The campus plan is a separate number
The release says Galaxy began operations in August 2026. It describes 3 gigawatts of planned capacity and support for up to one million AI accelerators at full build-out. Neither figure establishes how much capacity is installed or how many accelerators are running now.
The announcement provides no quantified deployed megawatt figure, utilization rate or measured cost saving. That limits what readers can conclude about the economics. We are not treating the company's broader efficiency claims as independently demonstrated results.
For investors following AI infrastructure, the useful follow-up would be operating capacity, measured reliability and a documented cost comparison. Those would help connect a technical deployment to an economic outcome. A large campus plan supplies context; it cannot stand in for that evidence.