China Mobile Spends $4.3B on Servers to Expand AI Infrastructure & Computing Power

China Mobile is massively scaling up its AI infrastructure with a US$4.3 billion order for servers, mostly from domestic manufacturers like ZTE, H3C, and xFusion. The intent is to build out intelligent computing centres and raise its national compute capability to 17 exaFLOPS, up from around 10-limit levels in certain segments.

The procurement includes two major tenders: one for approximately 265,000 servers, ordered at RMB 16 billion (≈US$2.2 B), and another for about 8,000 high-performance AI servers, costing RMB 15 billion (≈US$2.1 B). These will power various levels of computing platforms—national, provincial, and prefecture-level—to distribute AI compute closer to users and reduce latency.

One of China Mobile's newly commissioned intelligent computing centres, in Hohhot, Inner Mongolia, is capable of around 6.7 exaFLOPS. This is part of a wider rollout; twelve such centres have already been completed. Improvements include high-speed lossless networking, cold plate liquid cooling, and infrastructure enhancements to handle high server density.

China Mobile's capex allocation shows a shift: while overall capex is flattening or slightly falling, investment in computing infrastructure is up 21% year-on-year. As a share of capital, AI and intelligent computing now represent over a quarter of its total investment spend.

The server procurement also reflects China's localisation push: over 85% of the chips in this order are from local sources. Future tenders are likely to move even more toward domestically developed architectures rather than imported x86.

These developments position China Mobile not just as a telco, but increasingly as a telco-techco, with strength in both connectivity and compute. The upgraded infrastructure will underpin AI use cases across enterprise, citizen services, real-time analytics, 5G/6G, and beyond.