16:53Machine-translated from Azerbaijani by AIRead in Azerbaijani
Envision Energy Launches 800 V Power System for AI Data Center in China
“Envision Energy” has commercially deployed an integrated 800 V direct-current power system for a large artificial intelligence (AI) data center in China. The company calls it the world’s first such installation. The system has been undergoing testing since August 2026 at the “Galaxy AI” data center in Ulanqab, Inner Mongolia. According to the company, the setup is designed to power more computing equipment in less space.
In conventional data centers, electricity from the grid is converted several times before reaching the equipment. The “Dubhe” system converts 10 kV medium-voltage alternating current directly into 800 V direct current. It uses an electronically controlled solid-state transformer and a shared 800 V power bus. “Envision Energy” says the simpler design reduces energy losses and the size of the conversion equipment. The system also includes a battery energy storage unit that activates when the AI equipment’s power demand suddenly increases. The batteries provide additional power within seconds and serve as a backup source for several minutes. The transformer and energy storage unit are coordinated within a very short, millisecond-scale period. This balances sudden power fluctuations caused by graphics processing units (GPUs) within the data center. The “Galaxy AI” campus is planned to reach a capacity of 3 GW. The center will ultimately be able to support 1 million AI accelerators. The 800 V system allows server racks to be placed more densely. This enables more computing power to fit into the same space. The setup is designed to operate with renewable energy sources and energy storage systems. The “Dubhe Controller” uses artificial intelligence to align the power supply with computing demand. According to the company, this approach also reduces the need for cables and copper. “Envision Energy” plans to use the technology as the Ulanqab center expands. The system will also be used in the “Mission Gobi” project, which aims to build 5 GW of green AI data centers in desert regions by 2030.
In conventional data centers, electricity from the grid is converted several times before reaching the equipment. The “Dubhe” system converts 10 kV medium-voltage alternating current directly into 800 V direct current. It uses an electronically controlled solid-state transformer and a shared 800 V power bus. “Envision Energy” says the simpler design reduces energy losses and the size of the conversion equipment. The system also includes a battery energy storage unit that activates when the AI equipment’s power demand suddenly increases. The batteries provide additional power within seconds and serve as a backup source for several minutes. The transformer and energy storage unit are coordinated within a very short, millisecond-scale period. This balances sudden power fluctuations caused by graphics processing units (GPUs) within the data center. The “Galaxy AI” campus is planned to reach a capacity of 3 GW. The center will ultimately be able to support 1 million AI accelerators. The 800 V system allows server racks to be placed more densely. This enables more computing power to fit into the same space. The setup is designed to operate with renewable energy sources and energy storage systems. The “Dubhe Controller” uses artificial intelligence to align the power supply with computing demand. According to the company, this approach also reduces the need for cables and copper. “Envision Energy” plans to use the technology as the Ulanqab center expands. The system will also be used in the “Mission Gobi” project, which aims to build 5 GW of green AI data centers in desert regions by 2030.