Centrin Data's Yang Yu: Chips, not data centers, are AI's bottleneck
In an interview with China Entrepreneur magazine, Centrin Data executive Yang Yu said today's high compute prices come from a supply-demand mismatch rather than genuine scarcity, and that compute will eventually become as cheap and ubiquitous as electricity or water, with domestic chip progress—not data centers—the real bottleneck for AI. He said China has good power resources but lacks chips, while the US has chips but lacks power, and predicted leading third-party data center operators could reach multi-gigawatt scale by 2030. Yang also warned the sector will "compete itself to death" and that overcapacity may follow the AI boom, noting his company is investing early in Zhongwei, Horinger, Wuwei in Anhui and Changji in Xinjiang, while pursuing green-power direct connection in Ulanqab.
The interview offers an operator's view of China's compute supply-demand, chip bottleneck and site-selection cycle, useful context for gauging the timing of AI infrastructure investment.