A massive AI computing facility in Inner Mongolia is coming online, designed to accommodate up to one million accelerators and consume more than two gigawatts of power, signaling how the global race for AI infrastructure is now reshaping regions far from traditional tech hubs. Envision Group, a Chinese renewable energy company, has brought its Galaxy Campus online in Ulanqab, a city several hundred kilometers northwest of Beijing that is shifting from livestock and potato farming toward AI computing infrastructure.

Why Is a Pastureland City Becoming an AI Computing Hub?

Ulanqab's transformation reflects a practical reality about modern AI: these systems require enormous amounts of electricity, cooling, and physical space. The city's natural advantages make it attractive for large-scale data centers. Strong winds, long hours of sunlight, and cool climate conditions create ideal conditions for both renewable energy generation and the cooling systems that keep servers from overheating.

The Galaxy Campus itself is a sprawling facility. At its center sits a 120,000-square-meter AI data center, roughly equivalent to 20 standard football fields. The facility's planned power capacity exceeds two gigawatts, a figure that puts it in the same league as Colossus 2, the US-based AI data center operated by Elon Musk's SpaceXAI, which has a projected IT power capacity of 1.5 gigawatts.

However, important caveats apply. The figures cited are design and planning targets, not measurements of what is currently operating. The supplied report does not include independent confirmation of how many accelerators are actually installed, how much power is being consumed, or which companies and AI models will run on the infrastructure.

What Does This Mean for the Global AI Infrastructure Race?

The Ulanqab project illustrates a broader shift in how AI expansion is unfolding. As demand for computing power accelerates, the story is becoming as much about energy and infrastructure as it is about software and algorithms. Companies and regional governments are now competing to secure the physical resources needed to train and run increasingly large AI systems.

The combination of renewable energy and favorable climate conditions could offer practical benefits. Wind power may supply a portion of the campus's electricity, while cooler ambient temperatures could reduce the energy required for cooling systems. Yet the supplied report does not quantify these advantages or specify what portion of power will come from wind versus the grid.

How to Evaluate Claims About New AI Data Centers

• Planned vs. Operating Capacity: Distinguish between design figures and actual measurements. A facility announced as capable of housing one million accelerators may not yet have that many installed or operating at full power.
• Independent Verification: Look for regulatory filings, grid data, on-site reporting, or technical records that confirm a facility's actual status, rather than relying solely on company statements.
• Customer and Workload Details: Meaningful assessment requires knowing which AI models, companies, and applications will use the infrastructure, since training and inference workloads have different computing patterns and electricity demands.
• Environmental and Community Impact: Examine electricity mix, water requirements, emissions, land disturbance, and effects on surrounding communities, rather than assuming renewable energy automatically means low environmental impact.

The key questions for future reporting on the Ulanqab campus are straightforward: How many accelerators are actually installed? Is the full 120,000-square-meter facility active? How much of the planned power capacity has been energized? What customers and AI models will the campus serve? What is its actual electricity mix, water consumption, and environmental footprint ?

The comparison with Colossus 2 is instructive but incomplete. Both facilities represent major concentrations of computing resources if their plans are realized, yet accelerator counts and power capacity alone cannot establish how useful, productive, or cost-effective a data center will be. Two facilities with similar power budgets may deliver very different computing performance depending on hardware efficiency, cooling design, and software optimization.

For the public, the implications extend beyond AI companies. Large computing campuses affect electricity planning, transmission infrastructure, land use, water demand, and regional employment. In Ulanqab, those questions intersect with the pastoral economy that has historically defined the region. The report describes wind turbines standing beside livestock on surrounding grasslands, but provides no measurements or testimony about whether the new campus changes grazing conditions, local jobs, land values, or household electricity costs.

The Ulanqab project matters because it connects AI infrastructure requirements to a specific regional development strategy. It is a concrete example of how demand for AI capacity can reshape locations far from the companies that build or use the models. As AI systems become more powerful and more widely deployed, the infrastructure needed to support them will increasingly shape geography, energy policy, and regional economics around the world.