With too many idle servers, China may unleash a wave of discount CPU power like it did EVs
- The China Cloud Livestage Plan aims to sell the power leftover the processor of the inactive government data centers
- Despite massive investments, many Chinese data centers only operate by 20 to 30% capacity
- The old processors cost money even when it is inactive, China wants to monetize them before expiring
China moves its approach to the management of excess data center capacity by offering a new national system to redistribute the excess computing power.
After a three -year boom in the development of infrastructure, many local government centers are now facing low use and high operating costs.
As data centers are aging and fewer new customers need their services, the Chinese government aims to revive the viability of the sector thanks to a coordinated national service that would unify IT resources in all regions.
A coordinated response to increasing ineffectiveness
The proposal, motivated by the Ministry of Industry and Information Technologies (MIIT), consists in creating a network which allows the power of an excess processor of underused data centers to be grouped together and sold.
According to Chen Yili of China Academy of Information and Communications Technology, “everything will be given to our cloud to carry out a unified organization, orchestration and planning capacities”.
The objective is to provide a standardized interconnection of public calculation power on a national scale by 2028.
The overabundance emerged from the “Eastern Data, Western Computing” initiative, which has encouraged the construction of data centers in the less populated and energy -rich Western regions to serve the most developed oriental economic zones.
But many centers, despite the accommodation of some of the fastest processors, are now inactive, and it is a serious concern because the equipment in the data center has a specific lifespan.
In addition, processors and their related components are expensive to acquire and can become overwhelmed quickly, making unused infrastructure a financial liability.
Data centers are expensive to use and cooling systems, electricity and maintenance consume major resources.
Thus, when high performance workstation processors are left underused, they still incur current expenses, which is very bad for business.
The rates of use would have undertaken by 20% and 30%, undergoing both economic and energy efficiency.
More than 100 projects have been canceled in the last 18 months, a striking contrast with only 11 in 2023.
Despite the setbacks, the state’s investment remains substantial. Public markets reached 24.7 billion yuan ($ 3.4 billion) in only 2024, and 12.4 billion yuan have already been allocated in 2025.
The National Commission for Development and Reform (NDRC) has intervened to impose more strict controls.
New projects must comply with specific use thresholds and secure purchase agreements before approval.
In addition, local governments are now prohibited from launching small -scale computer infrastructure without clear economic justification.
On the technical level, the integration of CPUs of various manufacturers, including Nvidia and Huawei’s Ascend Chips, in a unified national cloud sets a serious obstacle.
The differences in hardware and software architecture make standardization difficult, and the initial objective of the latency of the latency of 20 milliseconds for real -time applications such as financial services is not satisfied in many distant facilities.
That said, Chen envisages a transparent experience where users can “specify their requirements, such as the quantity of computing power and the capacity of the network”, without worrying about the underlying chip architecture.
The question of whether this vision can be carried out depends on the resolution of infrastructure discrepancies and overcoming it the technical limitations which currently fragment the landscape of the calculation power of China.
Via Reuters
You might also love
