America has Stargate Abilene, the gigawatt-scale campus rising on the Texas plains for OpenAI. China has its own projects of similar size, and almost nobody outside the country has seen one up close. One of them—a gigawatt-scale campus in Inner Mongolia running mostly on wind and solar—is now offering a rare glimpse into the pace of China's AI infrastructure buildout.

Research firm SemiAnalysis got a look in September. The firm toured Envision Group's Galaxy AI datacenter in Ulanqab, Inner Mongolia, and called the speed, scale, ultra-low cost, and deep vertical integration it observed “astonishing.” First-hand reporting from inside a Chinese AI campus at this scale is uncommon, which is part of what makes the tour notable for observers tracking both countries' buildouts.

Inside the Galaxy campus

Envision, a major renewable energy company in China, started commercial operations at Galaxy in August 2026. The long-term target is 3 GW of capacity by the end of 2028.

For a sense of scale, the main site covers 120,000 square meters and is designed to hold up to 1 million AI accelerators—the specialized chips that train and run AI models. The anticipated computing capacity is one million PFLOPS. One PFLOPS means a quadrillion calculations every second, so the campus is built to run that workload a million times over. Power capacity, accelerator count, and computing throughput are the metrics used to size headline AI campuses on both sides of the Pacific, and Galaxy has set targets on all three.

The buildout is happening in phases. Roughly 120 MW has already been commissioned, including capacity tied to agreements with major Chinese tech firms. A 200 MW phase is expected to be handed over by late October 2026, and more handovers are planned into 2027.

Demand appears locked in well ahead of time. Envision has secured offtake agreements—long-term commitments to purchase the campus's capacity—running through 2029 to 2030, with customers in areas such as autonomous driving and content generation. That pre-committed demand gives the phased expansion a booked-in customer base as new capacity comes online.

Most of the power is meant to be green. More than 80% of the electricity is slated to come from wind and solar. The engineering also departs from the usual playbook: the facility uses what is described as the world's first commercial integrated 800V DC power system, purportedly designed for live AI workloads.

The “West Texas” of China

SemiAnalysis describes Inner Mongolia as China's fastest-growing AI datacenter region. It compares the area to West Texas, the American shorthand for wide open land, plenty of energy, and few neighbors to complain. The comparison ends at the politics: Texas is dealing with its highest-ever level of anti-datacenter sentiment, according to SemiAnalysis.

The region brings more than political goodwill. It has abundant renewable resources and a climate that helps with cooling, which matters when a building is packed with heat-producing chips.

Galaxy is the flagship of Envision's Mission Gobi initiative. That program aims to build 5 GW of green AI datacenter capacity worldwide by 2030. The math is notable: if Galaxy reaches its 3 GW goal, this single Ulanqab campus would make up the majority of the global Mission Gobi target.

As the phased buildout continues, the near-term markers are set: the 200 MW handover due by late October 2026, further handovers into 2027, and the 3 GW milestone targeted for the end of 2028—all steps on the way to Mission Gobi's 5 GW goal in 2030. Galaxy offers a rare public look at how quickly China is assembling AI infrastructure at gigawatt scale.

Source: CryptoBriefing