“Eastern Data, Western Compute” is Fake
ChinaTalk Irene Zhang
China's plan to shift AI compute westward mostly isn't happening. Most new data centers are landing in eastern exurbs, not the remote interior.
Based on reporting by ChinaTalk, Irene Zhang — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
China's much-hyped "Eastern Data, Western Compute" initiative was supposed to route data center buildout away from the crowded east coast toward the country's interior, where land, energy, and cooler climates are cheaper. It's a tidy story: Guizhou stores the data, coastal cities build the apps, and the whole country hums along as one integrated grid. Except a 2025 CAICT report on where China's chips actually sit tells a different story — Hebei, Guangdong, Jiangsu, Guizhou, Inner Mongolia, Shanghai, Zhejiang, Beijing, Shanxi, and Shandong top the list. Yes, Guizhou and Inner Mongolia show up. But this isn't a map of compute fleeing to the remote west — it's a map of a few data centers built out there to serve a thinner local population, while most capacity clusters near the coast.
The policy's origin story made sense on paper. Back in 2022, 80% of China's compute sat in the east while 70% of its renewable energy sat in the west, and officials like CAS engineer Zhao Xiaofang argued shifting facilities westward would both meet eastern demand and seed new digital industries inland. The NDRC leaned on the claim that electricity can eat up 70% of a data center's operating costs. But operating costs aren't the same as total costs, which are dominated by construction and chips — ChinaTalk's Aqib Zakaria estimates electricity is only about 5% of a 400 MW facility's three-year cost. Energy savings alone don't justify building far from population centers, especially once you factor in the skilled labor data centers need: electricians, HVAC techs, network engineers, cooling specialists. A 2021 report found 87.5% of China's data center talent concentrated in the north, east, and south — nowhere near the western hubs the slogan promised to build up.
What's actually happening looks less like a westward migration and more like sprawl into the exurbs of already-wealthy metros. Beijing bleeds into Hebei. The Yangtze River Delta spills into Jiangsu, Zhejiang, and Anhui. The Pearl River Delta spreads into cheaper corners of Guangdong. Even the NDRC's own 2022 map of eight "nexus points" for the buildout includes three eastern megacity clusters alongside four genuinely remote regions — Guizhou, Gansu, Ningxia, and Inner Mongolia — plus Chengdu-Chongqing, home to more than 50 million people. A March 2026 report from Communications Weekly, published under China's own industry ministry, flagged the predictable snags: western nodes stuck running general-purpose computing rather than AI-specific workloads, eastern firms reluctant to send workloads hundreds of miles away due to latency, and — the real bottleneck — chip shortages and a lack of high-end operations talent, largely due to semiconductor export controls.
The more uncomfortable part is what this has done to the western provinces that took the slogan at face value. China Economic Weekly reported in February 2025 that regions including Ningxia, Xinjiang, Guizhou, and Inner Mongolia rushed to build intelligent-compute parks on a build-it-and-they-will-come logic reminiscent of real estate speculation. Science and Technology Daily documented one western compute center, sited more than 20 kilometers from town for the cheap land and cool weather, running below 50% rack-up and under 30% actual server utilization, while burning through more than RMB 30 million a year in operating costs. By late 2024, China reportedly had nearly 150 such intelligent-compute projects running, with almost 400 more under construction or planned. That's a lot of concrete chasing a demand signal that may not show up.
The slogan sold poorer interior provinces on becoming genuine players in the AI economy rather than mere resource suppliers. Some, like Guizhou, may pull it off. But for many others, the data centers they've built on faith in a national compute grid look more likely to join China's long list of underused megaprojects than to deliver the AI boom they were promised.
My take — AI-written commentary, not fact-checked reporting
This is a familiar trap: a top-down slogan gets treated as a demand forecast, and provinces without the market signals to back it up end up holding empty server racks and multi-million-dollar power bills. Beijing's coordination can point capital in a direction, but it can't manufacture chip supply, skilled technicians, or actual customer workloads out of a policy document. The real losers here aren't the planners — it's the western regions that believed the pitch and built infrastructure nobody's using yet.”}}]}]}
Read more about this at: ChinaTalk