In March 2025 the Shanghai Municipal Commission of Economy and Informatisation issued its implementation opinion for 2025-2027 on promoting innovation in the intelligent computing cloud industry (Hu Jing Xin Ruan [2025] No. 160), setting a three-year development target for the sector.
Three target numbers
The opinion proposes that by 2027 the scale of the intelligent computing cloud industry should exceed RMB 200 billion; intelligent computing capacity should reach 200 EFLOPS, of which the autonomously controllable share should exceed 70 percent; and Shanghai should cultivate one to two strategic enterprises and 20 high-growth enterprises.
The fact that the “autonomously controllable compute share” is written into a quantitative target is what makes this document notable. It places the security attribute of compute supply alongside the scale target and points to the actual share of a domestic compute ecosystem.
From building compute to selling compute services
The phrase “intelligent computing cloud” means the centre of gravity shifts from physical compute construction to service-based supply: compute has to be delivered in cloud service form before small and medium-sized enterprises can actually use it. This dovetails with the scheduling logic of the action plan on compute interconnection: service-based supply presupposes that compute can be scheduled across entities.
See the national integrated compute network service platform.
Regional layout and industry cultivation
The document proposes deploying intelligent computing around key areas including Pudong, Jinshan, Songjiang, Lingang and Qingpu. A regional compute layout usually comes with arrangements for land, power and network resources, and it means the companies involved need to plan site selection and access in advance.
Three implications for companies
First, compute procurement needs to assess the autonomously controllable share. See our notes on the domestic trillion-parameter model and compute supply: guided by policy targets, the weight of domestic compute in the supply structure may keep rising.
Second, watch voucher-style instruments. See Shanghai’s AI application subsidy: support for compute cost is moving from direct subsidies to consumption and service vouchers.
Third, write compute supply stability into contract terms. See our notes on the open-source ecosystem and system solutions: a company that depends on a single compute supplier over the long term should agree on service interruption and migration paths.
Competition in compute policy will ultimately come down to who can provide more stable, cheaper and more controllable compute services.
This article is based on the Shanghai Municipal Commission of Economy and Informatisation’s implementation opinion on the intelligent computing cloud industry and public information and is for reference only; it does not constitute legal advice.