- China is rolling out a “computing power vouchers” program to help SMEs (small and medium-sized enterprises) access AI computing power at a low cost.
- The vouchers allow businesses to purchase services from national data centers at subsidized prices. Several cities have already adopted the program: Shanghai, Beijing, Shenzhen, Chengdu, Ningbo, Shandong, and Henan.
- Shanghai leads the way with a 600 million yuan ($84.1 million) budget for AI vouchers, which cover up to 80% of AI rental costs. The city also earmarked 100 million yuan ($14 million) for an LLM AI training data voucher program.
- Chengdu, which piloted the program in 2023, is expanding it with 100 million CNY for research institutes. Shandong has allocated 30 million CNY and is preparing an additional 1 billion CNY for AI infrastructure investment. Beijing and Henan have started accepting applications for support.
- The policy originates from a December 2024 document by the National Development and Reform Commission: to reduce R&D costs for SMEs and utilize surplus data center capacity.
- The “Eastern Data, Western Computing” strategy has led to the construction of hundreds of data centers in the west, where electricity is cheaper, to serve the industrial needs of the east. However, many centers operate at only 20–30% load, leading to overcapacity.
- China is building a unified national computing network to allocate loads among data centers and optimize utilization.
- The voucher program is expected to both reduce costs for businesses and address the infrastructure’s excess capacity, while also promoting domestic AI innovation.
- However, there are doubts about whether China can fully utilize this massive network, especially amid controversy over its plan to equip 39 new data centers with 115,000 smuggled Nvidia Hopper GPUs.
📌 China is accelerating its strategy to subsidize computing power for small and medium-sized enterprises, with Shanghai subsidizing up to 80% of AI rental costs with a budget of 600 million CNY (about $84.1 million). The program spans multiple provinces and cities, combined with a national computing network plan to leverage underutilized data centers in the west. This is a strategic move to expand domestic AI application but still faces questions about efficiency and long-term sustainability.
