Summary
- Envision’s AI Power System links renewable generation, grid interaction, storage, and computing loads.
- The architecture includes a 2.5MW solid-state transformer container converting 10/13.8kV AC to 800V DC.
- The launch reflects a wider move towards data centre power systems built around dynamic AI loads and grid scarcity.
Envision has launched integrated energy systems aimed at AI data centres, industrial users, and grids, placing power architecture closer to the centre of high-density compute design.
The company’s AI Power System combines renewable generation, grid interaction, energy storage, and computing loads. Envision says the architecture includes grid-side wind-solar-storage systems, medium-voltage grid-forming battery storage, and an 800V DC power chain using solid-state transformer technology.
The company’s Intersolar Europe material describes a 2.5MW solid-state transformer container that converts 10/13.8kV AC input to 800V DC output, with claimed conversion efficiency of up to 98.5% and copper-use reduction of up to 80%.
AI load pulls power upstream
GPU-heavy data centres are changing the electrical design brief. High rack densities, fast-changing load profiles, and large campus-scale demand are putting greater pressure on transformers, switchgear, UPS systems, batteries, busbar, controls, and grid interfaces.
Envision’s approach treats the data centre as part of an energy system rather than a passive connection. That direction is gaining ground as developers examine dedicated renewables, storage, microgrids, higher-voltage DC distribution, grid-forming inverters, and demand management to shorten delivery timelines and reduce exposure to constrained networks.
The 800V DC element points to one route for handling higher-power racks and reducing conversion losses. Higher-voltage DC architectures can support battery integration and improve efficiency, but they also introduce questions around protection, safety, maintenance, standards, and interoperability. Data centres adopt new electrical architectures slowly because uptime and serviceability carry more weight than theoretical gains.
Storage splits into two jobs
Envision has also referred to lithium iron phosphate storage for backup power and multi-hour energy shifting, and sodium-ion storage for fast smoothing and short-duration response. That combination separates two problems inside AI facilities: keeping the site running and managing rapid fluctuations.
Traditional critical-power design focused on resilience, transfer times, and backup duration. AI campuses add a grid-management problem. Large, dynamic loads can stress upstream infrastructure, and grid operators are increasingly sensitive to how major customers behave during peaks, faults, and constrained periods.
The market still needs proof from operating sites. Integrated power concepts must show reliable performance, clear failure modes, maintainable equipment, safety acceptance, and support coverage before conservative facility teams will adopt them at scale. Claims on conversion efficiency and copper reduction will be judged against whole-system reliability and cost.
Envision says its AI Power System is operating at its Chifeng Net-Zero Industrial Park and is linked to wider campus plans. The direct projects are outside Europe, but the design questions are already live across UK and EU data centre markets. Developers are searching for ways to bring compute online where grid access is slow, power quality is contested, or transmission capacity is scarce.
The launch underlines a shift in where AI data centre competition is being fought. Hardware still attracts the headlines, yet the electrical system increasingly decides when capacity can be delivered and how reliably it can operate.
