China Plans Aggressive Expansion Of Computing Power

China wants to boost its aggregate computing power by more than 50 percent by 2025, according to a plan released on Monday, as the country fights an increasingly bitter trade battle with the US focusing on areas such as high-end semiconductors and the specialised graphics chips used to train artificial intelligence tools.

The world’s second-biggest economy wants to have computing capacity equivalent to 300 exaflops  by 2025, according to the plan released by six key government ministries, including the Ministry of Industry and Information Technology (MIIT) and the powerful cyberspace regulator, the Cyberspace Administration of China (CAC).

The ministries said the increased computing power would be needed to support applications in areas such as finance and education.

In August officials said China had an aggregate of 197 exaflops of computing power, up from 180 exaflops in 2022, and said China’s average annual growth rate for computing power had reached nearly 30 percent over the past five years, with AI computing power growing by 45 percent annually.

Computing power

The officials said China now ranks second behind the US in computing power, without revealing their estimated figures for US capacity.

MIIT head Jin Zhuanglong said at the time that 41 percent of China’s economic output is generated by its digital economy, while the MIIT-affiliated China Academy of Information and Communications Technology projected that one yuan invested into boosting computing power would boost China’s gross domestic product by 3 to 4 yuan.

Jin said China was prioritising the expansion of its base of computing power as a key factor in driving toward economic self-sufficiency.

AI tools, which require vast computing power to train, have come into the spotlight this year and have reinforced China’s focus on building out its computing infrastructure.

Data network

Under the new plan China intends to build more data centres, particularly in less-populated western China, while improving the speed of data networks.

Transmission speeds between critical computing facilities must not have latencies of more than 5 milliseconds, the new plan says.

The US has been steadily imposing more trade restrictions on China to hobble its efforts to develop home-grown high end chips or to access the cutting-edge GPUs needed to train AIs.

The US, the Netherlands and Japan have all introduced significant export controls on high-end chips and chip-making technologies over the past year.