Envision has finished the first phase of its Galaxy AI data center campus in Inner Mongolia. The initial 120 MW of power was brought online in less than a year, according to a SemiAnalysis report visited by Tom’s Hardware in September. The company says the cost of this phase is about one‑fifth of what similar U.S. projects, such as the Stargate initiative, typically spend.
The campus is designed to eventually reach 800 MW, making it one of the largest AI‑focused data‑center clusters in the world. To keep the massive electricity demand affordable, Envision is building its own wind farms and battery storage next to the compute halls. This vertical integration of power generation and computing is unusual for data‑center operators, who normally buy electricity from the grid.
The move reflects China’s broader push to become a leader in AI infrastructure. By controlling both the hardware and the energy supply, Envision hopes to offer lower‑cost AI training and inference services to domestic firms. The project also aims to reduce reliance on imported power equipment, aligning with national goals of technology self‑sufficiency.
Why this matters
For AI startups and researchers in China, the cheaper power could translate into lower cloud‑compute bills, making large‑scale model training more accessible. For global competitors, the lower cost base may pressure U.S. data‑center providers to find similar energy‑savings strategies.
Why it matters
Chinese AI firms could run large models for less money, which may speed up development of new services. At the same time, the cost advantage could force U.S. providers to rethink how they power their own AI data centers.
