China Aims for 50% Increase in Computing Power to Compete with U.S.
China has announced an ambitious strategy to bolster its aggregate computing power by more than 50% by the year 2025, as disclosed in a recent release by government authorities. This move underscores China's heightened emphasis on supercomputing and advancements in artificial intelligence (AI).
This initiative arrives amidst escalating competition between China and the United States across numerous high-tech domains, encompassing semiconductors, supercomputers, and AI. Notably, the U.S. has imposed export controls on chipmaking equipment, intensifying the rivalry.
The comprehensive plan, jointly issued by six Beijing departments, including the Ministry of Industry and Information Technology (MIIT), sets an objective for China's cumulative computing power to reach 300 EFLOPS by 2025. EFLOPS, denoting one quintillion floating-point operations per second, gauges the computational speed of a computer.
As of August, the MIIT reported that China's computing power had surged to 197 EFLOPS in the current year, up from 180 EFLOPS in 2022. While the ministry acknowledged China's second-place ranking, it refrained from specifying the scale of U.S. computing power. Given the extensive computational demands of AI training, expanding the availability of computing power has become an increasingly focal point for Beijing.
According to a recent blog post by Google, top-tier generative AI models globally will necessitate tens of EFLOPs of AI supercomputing to sustain training durations of several weeks or less.
Under the outlined plan, China intends to establish additional data centers throughout the nation, streamlining access to computing power for businesses. Furthermore, to accommodate the burgeoning AI sector, Beijing intends to enhance computational infrastructure in the western regions of China.
Traditionally, less densely populated provinces like southwestern Guizhou have been entrusted with the establishment of extensive data centers to support the country's internet needs. Notably, Apple has collaborated with a local partner to establish data centers in Guizhou to cater to its user base in China.
Another pivotal objective is enhancing the speed and efficiency of the computational network. The plan stipulates that latency between critical computing facilities should not exceed 5 milliseconds.