China plans to quadruple AI computing capacity by 2030

Summary

China's Ministry of Industry and Information Technology announced a five-year plan aiming to expand artificial intelligence computing capacity from 2,185 exaflops to 9,800 exaflops by 2030. The plan includes 3.8 trillion yuan (approximately 488 billion euros) in infrastructure investment and promotes the use of domestic chips.

Why it matters

This is the most concrete signal yet that Beijing views AI computing infrastructure as a national security priority on par with energy or telecommunications. The explicit prioritization of domestic accelerators signals a hard shift toward technological self-sufficiency under US export restrictions. The massive scale, featuring clusters of over 100,000 GPUs and dedicated inference data centers, would fundamentally reshape the global AI supply chain and competitive landscape if realized.

Highlights

• Current capacity has increased by 177% year-on-year, with each of the 52 data centers hosting over 10,000 accelerators.
• According to the China National Data Administration, daily token consumption rose from approximately 100 billion in early 2024 to about 140 trillion by March 2026.
• The plan builds on the 'East Data, West Computing' initiative launched in 2022, which aims to locate new data centers in lower-cost western regions.

Critical take

The article presents the 1,400-fold increase in daily token processing volume as a simple fact, without questioning the measurement methodology or the consistency of token counting across platforms. From a broader perspective, the text approaches China's AI momentum almost like a hagiography. It fails to discuss whether domestic chips can actually sustain 9,800 exaflops of capacity at the required efficiency. Additionally, it does not address the energy and cooling constraints in the western regions where new clusters are being established.