China’s dark compute power could be 6,000 times higher than current estimates

AdvertisementScienceChinaScienceChina’s dark compute power could be 6,000 times higher than current estimates

Domestic AI amounts to 1,882,000,000,000,000,000 calculations per second, vastly more than Western rankings suggest

2-MIN READ2-MIN2 Listen

Ling Xinin OhioPublished: 9:00am, 23 Apr 2026China has reported a staggering amount of domestic artificial intelligence power, pointing to what some experts say could be a “dark pool” of compute thousands of times larger than its public reports suggest.

According to the Ministry of Industry and Information Technology (MIIT), China has achieved 1,882 exaflops – short for exa floating-point operations per second – which translates to 1,882 quintillion, or billion billion, calculations per second.

The figure is more than 6,000 times higher than the country’s computing power as reflected in the Top500 list, one of the few available benchmarks for comparing China’s supercomputing progress with other countries, especially the United States.Advertisement

While the two are not directly comparable, and the Germany-based ranking of the world’s fastest supercomputers is often seen as an underestimate, the latest data released by the MIIT on Tuesday suggests China’s computing power is rising fast.

Unlike China, the US – which dominates the Top500 – does not publish a single national figure for AI computing power as most of the infrastructure is owned by private companies and measured using different standards.

01:46

AI rush turns everyday data storage into ‘digital gold’ for Hong Kong consumers

AI rush turns everyday data storage into ‘digital gold’ for Hong Kong consumers

According to estimates from the Stanford Institute for Human-Centred Artificial Intelligence and other policy analyses, the US holds roughly 50 to 75 per cent of global capacity and hosts the largest concentration of AI-focused data centres.

AdvertisementAdvertisementSelect VoiceSelect Speed0.8x0.9x1.0x1.1x1.2x1.5x1.75x00:0000:001.00x