Shanghai said at the 2026 Global Investment Promotion Conference on March 14 that it will launch 31 “new quality factors” to attract investment, with AI at the center of the package. The city highlighted a plan to build what it calls China’s largest computing power dispatching platform, offer “use first, pay later” access to compute, and issue 1 billion yuan in annual compute vouchers with “no application required.” The package also includes 11 public service platforms, 10 specialized pilot (test) platforms, and 10 benchmark application scenarios, signaling a city‑level move to make AI infrastructure more accessible and cheaper.
Official and media coverage framed the 31 measures as a bundled release rather than a single policy. Xinhua and China News Service reported that the package mixes infrastructure with real‑world scenarios, a sign that Shanghai wants to match supply (compute, platforms) with demand (industry applications). By bundling 11 public service platforms with 10 pilot platforms and 10 benchmark scenarios, the city is effectively promising end‑to‑end support: foundational services, testing and validation, and deployment environments that can absorb AI solutions.
The compute measures are the most concrete and quantifiable. Securities Times and Eastmoney cited the city’s plan to build the largest compute dispatching platform in China, designed to coordinate access to roughly 140,000 P of heterogeneous compute capacity across Shanghai. The same reports said enterprises will be able to access compute on a “use now, pay later” basis and claim 1 billion yuan in annual vouchers with “no application required,” a structure meant to remove administrative friction and reduce the cash‑flow burden for AI teams.
The platform‑plus‑voucher approach is a notable policy design. A dispatching platform suggests a city‑level attempt to pool compute resources across multiple providers and make capacity schedulable, rather than leaving companies to negotiate fragmented access on their own. Pairing that with vouchers and deferred payment terms is effectively an operating subsidy. It turns compute from a fixed upfront cost into a variable cost supported by public funds, which matters when training or fine‑tuning models can require large bursts of GPU time.
Shanghai has been signaling that AI is a core growth engine, echoing broader policy signals such as the national AI legislation push, and recent industry figures provide the backdrop. The Shanghai commission for economy and informatization has said the city’s AI industry scale reached roughly 6,300–6,370 billion yuan in 2025, with growth around 39.5% year on year, and 394 firms classified as above‑scale enterprises. Public data also cited 149 generative‑AI service filings in the city, indicating a broad base of companies already building or deploying generative products.
Those numbers help explain the political logic of a compute subsidy. If a city is already supporting hundreds of sizeable AI firms and a fast‑growing generative‑AI cohort, the marginal value of cheaper compute is high. A 1 billion yuan voucher pool could meaningfully affect the training budgets of startups and mid‑sized firms, particularly those without long‑term GPU supply contracts. In other words, Shanghai is trying to turn compute from a bottleneck into a public utility that can accelerate commercialization.
The 31‑item package also reflects a broader shift in local industrial policy: the focus is no longer just on attracting headquarters or offering generic tax breaks, but on providing tangible production factors. The breakdown—11 public service platforms, 10 pilot platforms, 10 application scenarios—reads like a pipeline for AI products, from infrastructure to testing to deployment. That matters because AI firms often struggle not only with compute costs but also with access to real‑world data, regulated environments, and early adoption by public or industrial partners.
There are still operational questions that will shape the real impact. The size of the compute pool (140,000 P) is large, but how much of it is actually schedulable for third‑party developers, and under what pricing tiers, remains to be seen. The “no application required” voucher language suggests a frictionless claim mechanism, but it raises questions about eligibility rules, caps per company, and how the city will prevent over‑consumption or misuse. The exact design of those rules will determine whether the policy becomes a broad accelerator or a narrow benefit for a small set of firms.
From a national competition angle, Shanghai’s move is also a signal to other Chinese cities, including recent district‑level packages like Shenzhen’s Luohu AI+OPC support plan: if a global financial and industrial hub is willing to subsidize compute at scale, smaller cities may need to respond with either targeted subsidies or specialized vertical programs. The policy could therefore set a benchmark for how local governments define “new quality productive forces” in the AI era, shifting from slogans to specific infrastructure and voucher programs.
What changed is that Shanghai elevated AI compute from a firm‑level procurement problem to a city‑level public service, with a defined voucher pool and a dispatching platform tied to 31 new investment factors. What may happen next is a race among cities to offer comparable compute support and application scenarios—and, inside Shanghai, a rapid test of whether the voucher mechanism and platform governance can translate into faster model deployment and measurable industrial growth.
Sources
- http://www.xinhuanet.com/fortune/20260314/eab714205c74440794b5a4752e309261/c.html
- https://www.chinanews.com.cn/cj/2026/03-14/10586733.shtml
- https://www.stcn.com/article/detail/3677341.html
- https://finance.eastmoney.com/a/202603143672241337.html
- http://sh.people.com.cn/n2/2026/0307/c134768-41517546.html
