China’s AI Spending Shifts to Inference, Telecom Operators Lead the Build
RMB 51 billion in inference servers, 222 tenders, and a procurement trail that shows who controls the capacity.
Editor’s Note:
This is FlashPoint, Hello China Tech’s premium quick-strike column on one market-moving China tech event and why it matters.
Today’s column examines China’s AI infrastructure buildout: RMB 51 billion in inference servers, 222 procurement projects, and a shift in what kind of capacity is being prioritized. The key question is who is financing this shift, and what it means for the AI stack.
China’s AI buildout is entering the “10,000P era,” shorthand for massive compute clusters. The real question is what kind of capacity is being built, and who controls the cheque book. Early signals point in one direction: inference is rising to the top of the build list, and the biggest cheques are coming from the national carriers.
Public tender activity in 2025 provides a rare window into the answer. A compilation drawn from official procurement and bidding platforms, plus corporate database records, identified more than 222 data-center and intelligent-compute projects with individual winning bids above RMB 100 million. These covered everything from civil construction and power infrastructure to IT architecture and server purchasing. They were spread across 29 provinces and more than 87 prefecture-level cities. The scale ceiling rose again: at least 37 projects cleared RMB 1 billion, and many public compute centres moved from “hundreds of P” to “thousands of P” and into the “10,000P-plus” range.
Two shifts stand out.
