The digital possibilities of frontier AI depend on the expansion of physical data centers.
These massive data centers require contiguous land, high-density power connections, and massive cooling systems that compete directly with local agriculture and municipal resources. To ease this resource burden on populous coastal regions, China’s Eastern Data, Western Computing (EDWC) initiative aims to relocate and expand data centers in western China—taking advantage of greater land availability, lower electricity costs, renewable-resource endowments, and cooler climate conditions.
What geographic and infrastructure criteria is China using to plan this AI data center expansion? Where in western China are the ideal regions for building out these data centers, and why? How should policymakers view and respond to this geographic rebalancing of China’s data center strategy?
To explore questions like these and more, this NSPDI paper analyzes China’s EDWC initiative using multi-source geospatial data, remote sensing inputs, and Pareto-based spatial screening. In this study, we argue that governments must manage AI as a strategic resource system, bounded by a material trilemma of resource availability.
Key Takeaways:
- China’s western hubs each face trade-offs for data center expansion, as no hub simultaneously maximizes grid accessibility, water security, and land suitability. Ningxia is the only hub where all constraints align favorably at the site level.
- Water is a hidden binding constraint on China’s AI infrastructure ambitions. High-density AI compute requires continuous liquid cooling, and the sites with the strongest renewable access consistently face elevated water stress.
- EDWC is a national security infrastructure project, not just a digital economy policy. By concentrating compute, power, and cooling capacity in western hubs insulated from foreign hardware chokeholds, China is building a domestic AI production base designed to be resilient to U.S. export controls.