Envision says its AI campus’s first phase is booked, with over 80% direct wind power planned

Envision says the first phase of its Galaxy Campus AI data center in Ulanqab, China, is fully booked. More than 80% of the facility’s electricity will come directly from its wind farms, the company says. CNN’s October 13 reporting details the under-construction campus’s customer commitments and energy design, while advanced-chip access remains a challenge.

The two-gigawatt figure describes the planned campus as a whole; the booking disclosure covers only its first phase, not two gigawatts of installed computing capacity. General manager Ricky Zheng said Envision was building additional phases for AI and cloud services.

Customers own the AI chips and servers, Zheng told CNN. He said their construction orders reflected confidence that they could supply enough hardware to fill the racks. The campus is intended to handle workloads including questions submitted to AI apps by users in places such as Beijing.

Wind power without the grid detour

Envision’s planned direct wind supply would bypass the grid for most of the facility’s electricity. Turbines around 100 miles from Galaxy Campus contribute power directly to the site. Zheng said batteries store wind-generated electricity, while a solid-state transformer—equipment that adjusts electrical power—responds to changing server demand.

Plentiful power, constrained chips

Galaxy Campus sits within a wider buildout: nearly 90 data centers are slated for Ulanqab, a city of fewer than two million people. The location fits China’s Eastern Data, Western Compute strategy, which directs computing infrastructure toward regions with plentiful land and renewable energy.

Hardware remains harder to secure. US export restrictions limit Chinese firms’ access to advanced Nvidia and AMD AI chips. Domestic chipmakers still trail US rivals, and supplies of their cutting-edge models are tight, CNN reports.

Abundant electricity offers a partial workaround. Chinese chip designers such as Huawei combine chips into clusters to approach the computing power of more capable hardware. Analysts cited by CNN say this can work for some tasks, but uses significantly more electricity.