China to boost its state-owned compute performance by 30% by 2025, plans to hit 300 exaflops

China is set to increase its national computing power by 30% by 2025, growing from 230 ExaFLOPS to 300 ExaFLOPS, reports The Register. This ambitious plan involves significant advancements in technology, though for now it leaves more questions than answers as gaining an additional 70 ExaFLOPS of compute power is not an easy thing to do.

Currently, China has over 8.1 million data center racks providing 230 ExaFLOPS of processing power, according to data revealed at the Global Digital Economy Conference 2024 by Wang Xiaoli from the China Academy of Information and Communications Technology. 

China's plan is to reach 300 ExaFLOPS by 2025, which is a considerable increase when we look back over previous growth rates. China's compute power grew from 180 ExaFLOPS in 2022 to 197 ExaFLOPS in August 2023, which means that the new goal will require a significant acceleration. Unfortunately, Wang Xiaoli did not say how China is going to add 70 ExaFLOPS of compute horsepower in less than a year from now.  

How China plans to add 70 ExaFLOPS of compute capacity amid sanctions by the U.S. that require companies like AMD, Intel, and Nvidia as well as their partners to get an export license to sell their high-performance processors or servers to Chinese entities, both public and private, is a big question. The country can barely produce its own high-performance processors, as contract chipmakers like Semiconductor Manufacturing International Co. (SMIC) cannot procure advanced wafer fab tools as well. 

China's plan to enhance its computing capacity aims to support broader AI deployment and economic transformation, particularly in rural areas, which means that the plan is important for the whole country. This increase in compute power is expected to drive significant technological and economic benefits across China, but, again, if China manages to get appropriate hardware. 

But while it is problematic for Chinese companies to get high-performance computing processors and servers, they can get energy storage technologies. Yovole Network, a Shanghai-based cloud computing data center service provider, has partnered with Tesla to implement its Megapack energy storage technology. Yovole Network is also incorporating advanced technologies such as hydrogen energy, photovoltaic storage, indirect evaporative cooling, and liquid cooling in its data centers, which indicates that it has some massive compute horsepower inside. These technologies are essential for supporting the increased processing power while managing energy efficiency.