Overview

Envision announced on Nov. 2, 2026, that its Galaxy AI data center campus in Ulanqab, Inner Mongolia, China, has completed end-to-end commissioning of an integrated 800V DC power system and is now powering live computing workloads. The company describes the installation as the world's first commercial deployment of such an architecture in a large-scale AI data center. The system delivers power from grid input to computing loads, combining solid-state transformers, battery energy storage, DC/DC converters and AI-driven controls. The 3GW campus is positioned as a commercial reference for the industry's move to 800V DC infrastructure.

Technical Architecture

At the center of the deployment is an integrated power chain that converts 10 kV medium-voltage AC input directly to an 800V DC bus across operating conditions. Envision's solid-state transformers perform the conversion, while battery energy storage provides both second-scale power support and minute-scale backup. DC/DC converters adjust direct-current voltage for computing hardware. The Dubhe Controller, supported by the Dubhe Energy Foundation Model, coordinates transformers and batteries in real time to match power supply with computing demand and manage sudden load swings.

The architecture is notable for its millisecond-level coordination between SST and storage. Instead of relying solely on upstream grid response, the system absorbs GPU load fluctuations locally. This is important because AI workloads can swing sharply from moment to moment, and operators must keep supply stable while delivering more power to each rack within constrained space.

Why 800V DC Matters for AI Racks

Envision says the 800V DC architecture simplifies AI data center power systems, bringing higher power conversion and watt-to-bit efficiency, a more compact equipment footprint, and reduced cabling and copper requirements. Those advantages can lower infrastructure costs and support higher computing density within the same space. The company says the architecture supports higher-density AI racks, enabling more computing power within the same footprint.

The shift to 800V DC is emerging as an industry response to accelerating AI computing demand. Higher-density racks require more power delivery capacity, but space and cooling constraints limit conventional distribution approaches. By delivering a DC bus closer to computing hardware, the integrated system aims to reduce conversion complexity and physical infrastructure while maintaining stability under dynamic loads.

Commercial and Operational Significance

The deployment is not a pilot or laboratory demonstration. According to Envision, the system is now powering live computing workloads, making it a commercial reference ahead of wider industry adoption. The Galaxy campus began operations in August 2026. With a planned capacity of 3GW, it is designed to support up to one million AI accelerators operating in parallel at full build-out, with more than 80% of its power planned to come from direct renewable electricity connections.

Kevin Huang, Senior Vice President of Envision Energy and President of Energy Storage and AIDC Power Systems, said: 'AI data centers need power systems that coordinate renewable energy, energy storage, and the grid to meet computing demand. Our integrated 800V DC system is now powering live computing workloads, demonstrating an end-to-end architecture that connects renewable energy supply with computing demand.'

Market Context

The announcement comes as the industry advances its transition to 800V DC infrastructure for next-generation AI data centers. Envision's integrated system is operating commercially ahead of wider adoption, providing a reference for that transition. The market context is defined by three pressures: rising power demand per rack, constrained space, and the need to keep supply stable as computing loads swing sharply. At the same time, operators are seeking to coordinate renewable energy, storage and grid interaction more tightly.

Envision aims to scale the approach through its Mission Gobi initiative, which targets 5GW of green AI data center capacity in desert regions worldwide by 2030. The Galaxy campus and the Mission Gobi target together frame the company's strategy: pair large-scale renewable electricity with integrated power electronics and storage, then use AI-driven controls to align supply with computing demand.

Implications for AI Data Center Design

For AI data center designers, the deployment suggests that the power chain is becoming a more integrated part of the computing platform. Combining SSTs, battery storage, DC/DC conversion and real-time control in one architecture changes the boundary between facility infrastructure and IT load. Local absorption of GPU fluctuations can reduce the burden on upstream systems, while a simplified DC distribution path can reduce cabling, copper and equipment footprint.

The commercial reference also highlights remaining industry questions. 800V DC adoption will require standards, protection schemes, safety practices and operational experience across the ecosystem. Envision's deployment provides one data point: an integrated, vendor-coordinated system operating at scale in a live AI campus. Whether the approach becomes a broader template will depend on how operators weigh efficiency and density gains against integration risk and supply-chain readiness.

Future Outlook

Envision's 3GW Galaxy campus and its 5GW Mission Gobi target indicate that the company intends to scale integrated 800V DC power systems beyond a single site. If the architecture proves repeatable, it could accelerate the industry's shift away from conventional AC distribution in high-density AI facilities. The key test will be whether the system's promised efficiency, footprint and coordination benefits hold across diverse workloads and renewable supply conditions.

For now, Envision has moved the 800V DC concept from industry discussion into commercial operation. The deployment provides a concrete reference for power-compute coordination, linking renewable energy supply, battery storage and AI computing demand through a single integrated power system. Its significance lies less in a single component than in the end-to-end architecture now operating at a large-scale AI data center.