- Gongshu bets on open-source models while Qianjiang Century City pushes data infrastructure
- Citywide AI expansion deepens competition across chips, models, data and applications
Hangzhou is entering a new phase in its push to become China’s leading AI hub, with districts across the city launching competing initiatives spanning open-source models, AI data infrastructure and embodied intelligence.
On May 21, Hangzhou’s Gongshu District unveiled what it called a “DeepSeek-plus” open-source ecosystem initiative alongside a package of supporting policies.
Just days earlier, on May 18, the ZODA international open-data community was officially launched in Qianjiang Century City—one of the city’s central business districts—marking another major AI-related infrastructure project in the city.
The moves come as Hangzhou intensifies efforts at the start of its “15th Five-Year Plan” period (2026-2030) to position itself as what officials call China’s “top city for AI innovation and development.”
As the registered home of DeepSeek, Gongshu District is seeking to leverage its early advantage in open-source large models.
Its newly released action plan outlines a strategy centered on expanding open-source AI applications through initiatives including model ecosystem expansion, domestic computing infrastructure and industry-specific deployment scenarios.
The district said it has already attracted or incubated 125 companies tied to the “DeepSeek plus” ecosystem.
Local firms including Runda Medical and Zhejiang Digital Culture and Tourism have integrated DeepSeek models into medical consultation and tourism-assistant services.
Gongshu also opened a new offline developer hub called “Canal Origins · DeepSeek Home,” intended to serve as a gathering space for global developers and lower entry barriers for smaller AI startups.
While Gongshu focuses on models and applications, Qianjiang Century City is targeting a different bottleneck in the AI supply chain: data.
The newly launched ZODA community aims to tackle shortages in high-quality datasets and standardized model evaluation systems.
Organizers said more than 80% of AI development time is currently spent on data preparation, making data quality one of the key constraints on model performance.
The project plans to build open-source datasets, establish model benchmarking standards and connect international expert networks. It also intends to create an offline “ZODA House” modeled after Silicon Valley-style developer communities.
Organizers said the initiative aims to drive more than 10 billion yuan ($1.47 billion) in AI data industry activity within two years and eventually establish “Hangzhou data standards” as a reference point for global AI developers.
Other districts across Hangzhou are also accelerating AI-related development. In the city’s Yungu area, China’s largest AI open-source community, ModelScope, has launched an offline developer center designed to attract global AI developers.
At the city level, Hangzhou is also rapidly building out an embodied AI industrial chain and recently introduced China’s first local regulation specifically targeting embodied intelligence robotics.
Taken together, the initiatives reflect how Hangzhou is attempting to assemble a full-stack AI ecosystem spanning chips and computing power, data infrastructure, open-source models and downstream applications, with different districts carving out specialized roles across the broader industry chain.
