Unveiling "Western Shrimp, Eastern Farming": Three Logics Behind OpenClaw's Localization Explosion

An open-source AI agent software, OpenClaw, has sparked a nationwide “lobster farming” craze in China. Nearly a thousand people queued outside Tencent’s headquarters in Shenzhen to install it, including programmers, retired engineers, housewives, and even elderly people actively participating.

The speed and enthusiasm with which Chinese people embrace AI have left foreigners stunned.

“Quite amazing!” wrote Philip Sun, Asia Sales at Goldman Sachs, in a memo. “The young Chinese investors I’ve interacted with already have several ‘lobsters’… One summarizes market news every morning, another analyzes whether investment decisions are too emotional, and there’s even a ‘supervisor lobster’ watching over the others.”

(Tencent employees set up a booth outside the Shenzhen headquarters, offering free OpenClaw installation assistance to passersby.)

This wave of “AI lobster” fever not only shocks Wall Street but also triggers a deep reassessment of the industry logic behind the “Western shrimp, Eastern养” (farming) model. On March 11, Guolian Minsheng Securities released a research report titled “Western Shrimp, Eastern养—A Comparison of OpenClaw’s Development in China and the US,” analyzing the underlying business logic of this trend.

“Western Shrimp, Eastern养”: The Chinese Logic of Industry Implementation

Although OpenClaw originated overseas, after February, the Chinese market quickly adopted and scaled its application, exhibiting a unique “Western shrimp, Eastern养” characteristic.

The report suggests that this wave of “Western shrimp, Eastern养” is mainly driven by domestic large companies’ “FOMO-style anxiety” over OpenClaw’s market trend. In reality, it reflects an ecological mismatch in the AI application deployment wave between domestic and overseas markets during the Agent era. Compared to overseas, China has advantages in cost, traffic, and policy.

From a cost perspective, China’s large model ecosystem has formed a unique “low-cost API” barrier. The report notes: “The API call price for domestic models is about one-sixth of similar overseas products.” This advantage stems from cheaper electricity behind domestic computing power, more flexible hardware configurations, and fierce competition among model vendors.

Moreover, driven by OpenClaw, domestic model token consumption has accelerated. Data from OpenRouter shows recent record highs in large model token usage, with steep growth rates, and Chinese models rank among the top in token consumption. The rapid increase in token usage also proves the cost advantage.

From a traffic perspective, active participation by major companies is key to the “lobster fever” explosion. With Zhipu launching “AutoClaw” on March 10, domestic “lobster products” have expanded deployment methods to include local, cloud, and hybrid setups.

For internet giants, deploying OpenClaw may not only boost revenue from Agent-related services but also involve competing for potential “new distribution rights” in the Agent era. Looking ahead, as Agent deployment continues, interactions between humans and the digital world may undergo a fundamental restructuring—when the primary operator shifts from humans to Agents, new traffic entry points may also migrate.

Although this wave of OpenClaw may still face technical immaturity, its shift of AI from “Chat” to “Work” could trigger market concerns about changes in new platform entry points, leading major companies to experience FOMO and actively deploy. Additionally, giants like Tencent, leveraging social fission effects, could further boost traffic attention, spreading societal focus on the “lobster craze.”

From a policy perspective, regions like Shenzhen and Wuxi have already introduced special policies. The report mentions that Wuxi High-tech Zone issued 12 “lobster养” policies, with individual support up to 5 million RMB, covering talent recruitment, industry implementation, safety, and compliance.

1. On March 8, Shenzhen announced that the Longgang District AI (robotics) bureau is soliciting opinions on the “Support Measures for OpenClaw & OPC Development (Draft for Comments).” It encourages market-based, professional platforms to launch “lobster service zones,” offering free deployment services for OpenClaw, with subsidies for qualified entities.

2. On March 9, Wuxi High-tech Zone released “Measures to Support the Integration of OpenClaw and Other Open Source Community Projects with OPC Community.” The 12 policies range from basic support to industry implementation, talent recruitment, and safety compliance, offering a package of financial incentives, with individual support up to 5 million RMB. For deployment, local cloud platforms providing free deployment and development kits can receive full subsidies up to 1 million RMB.

Localization of Deployment Forms and Model Selection

The latest development in this wave is the active involvement of leading large model vendors.

Guolian Minsheng Securities reports: “On March 10, Zhipu officially launched AutoClaw (Chinese name: 澳龙), a one-click installable local version of OpenClaw with full native capabilities.”

According to Zhipu, “AutoClaw 澳龙 comes preloaded with over 50 mainstream Skills, ready to use out of the box, covering high-frequency scenarios like content creation, office work, coding, marketing, and financial research, supporting quick integration with tools like Feishu.”

Not only Zhipu, but domestic giants are also fully participating in localization. Since February, companies like Zhipu and MiniMax, along with major cloud providers, have released related Claw products.

Guolian Minsheng Securities observes that domestic products now show distinct features in deployment methods and model sources.

In deployment, the report states: “From the perspective of deployment methods, the current wave of ‘lobster’ participation is highly diverse.” Unlike overseas OpenClaw, which emphasizes local deployment, domestic “lobster products” feature local, cloud, and hybrid deployment options, offering greater flexibility. The report attributes this mainly to domestic “lobster products” focusing more on deployment convenience and security.

Regarding model integration, the report emphasizes: “Currently, most domestic entities support multi-model matching, but some choose to rely solely on their own models.” Compared to the broader support in overseas versions, domestic “lobster products” mainly support calls to mainstream domestic large models, with some only supporting their own models as primary sources.

Overall, the report believes that under the influence of this wave of OpenClaw-driven “lobster fever,” China’s AI industry continues its trend of application deployment, demand acceleration, and industry chain prosperity.