"Lobster" Frenzy: A Half-Month Chronicle - Shenzhen Policy Support, Concept Stocks Surge and Plunge, Is AI Agent a Windfall or Bubble?

Chinatimes.net.cn Reporter Hu Mengran, Shenzhen

“Have you been raising lobsters lately?”

In early 2026, this topic broke out of the aquaculture circle and became a buzzword in the tech community. The “lobster” here doesn’t refer to seafood but is the name of an open-source AI agent called OpenClaw, named after its red cartoon lobster icon. The process of deploying and training this AI agent is vividly called “raising lobsters.”

Since the first official version was released at the end of January, this wave driven by open-source technology has rapidly spread among tech enthusiasts: Shenzhen Longgang quickly introduced the “Lobster Ten Policies” support measures, regions across the country launched “raising lobsters” experiences, major companies competed to offer “cloud lobster” services, the A-share market responded swiftly, and concept stocks surged.

However, as the hype grows, the tide is turning.

When authority is handed over to “lobsters,” issues such as privacy leaks, file loss, plugin poisoning, and other security risks also emerge. On March 11, the Cybersecurity Threats and Vulnerabilities Information Sharing Platform of the Ministry of Industry and Information Technology issued the “Six Do’s and Six Don’ts” advice on preventing security risks of OpenClaw (“lobster”). After continuous gains, concept stocks experienced sharp corrections, and doubts about how long the “information gap dividend” can last are rising.

Shenzhen Releases “Lobster Ten Policies”

The “Support Measures for the Development of OpenClaw & OPC in Longgang District, Shenzhen (Draft for Comments)” (hereinafter referred to as “Lobster Ten Policies”) is currently open for public consultation, with the comment period from March 7, 2026, to April 6, 2026.

The “Lobster Ten Policies” propose substantial support: offering up to 2 million yuan in technical research subsidies and 10 million yuan in equity investment support to global AI developers and “one-person companies” (OPC) entrepreneurs.

The quick issuance of special policies within a short period of OpenClaw’s popularity is astonishing.

“Longgang District’s rapid response reflects the local government’s high sensitivity and proactive attitude toward emerging technologies, representing a swift response to the national strategy of ‘cultivating new forms of intelligent economy,’” said Huo Hongyi, director of the Asia-Pacific Social Innovation Research Institute, in an interview with Huaxia Times. He pointed out that this indicates a profound change in local investment attraction strategies: shifting from past focus on large enterprises and broad industry layouts to focusing on emerging niche fields and cultivating characteristic industrial ecosystems.

Jiang Han, senior researcher at Pangu Think Tank, described this phenomenon as “the AI version of Shenzhen speed.” He told Huaxia Times that competition among districts in Shenzhen is fierce. Longgang’s support for OpenClaw and OPC aims to create a differentiated competitive edge and seek new economic growth points. This reflects a shift from “investment attraction and intelligence gathering” to “ecosystem cultivation,” from “hardware competition” to “service competition.” Instead of merely attracting large enterprises, the focus is on nurturing the industrial soil by subsidizing computing power, data, and other elements to support emerging startups and foster an entrepreneurial environment.

The “Lobster Ten Policies” also tailor support plans for the new entrepreneurial model of “one-person companies”: newly settled enterprises can receive up to three months of free computing power, and qualified projects can get up to 4 million yuan in scene support.

What conditions are needed for “super individuals” to grow into sustainable market entities besides computing power and space? “The most urgent factors to break through are market channels and business operation capabilities,” Huo Hongyi pointed out. “Super individuals often have strong technical skills but may lack experience in market promotion, customer acquisition, and business model development. Building and improving talent teams is also crucial—initially, a one-person company may be led by an individual, but as the business develops, it needs to introduce professionals from different fields.”

Jiang Han added three key elements: talent, market channels, and continuous funding support. He said, “Super individuals need to open up markets, establish customer bases, and without resource and channel accumulation, sustainable operation is difficult. Although there are subsidies for computing power, business expansion and R&D still require funds. Besides government equity investments, more social capital needs to be connected.”

When the “key” is handed over to AI, who will guard the bottom line?

When authority is handed over to “lobsters,” issues like privacy leaks, file loss, plugin poisoning, and other security risks also follow. Observers note that users excitedly “raise lobsters” while also paying to uninstall them.

On March 11, the Cybersecurity Threats and Vulnerabilities Information Sharing Platform issued the “Six Do’s and Six Don’ts” advice on preventing security risks of OpenClaw. On March 10, the National Cybersecurity Emergency Response Center issued a risk alert titled “Risks of OpenClaw Security Application,” pointing out that while “lobster” has gained widespread attention for supporting natural language control of computers, its high system permissions and weak default security configurations have exposed serious vulnerabilities. Attackers can exploit these flaws to easily gain full control of the system. The core risks identified include prompt injection, misoperation, plugin poisoning, and severe security loopholes.

Given the high risks of “handing the keys” to AI with extensive permissions, can current technical defenses keep pace with the rapid adoption?

Guo Tao, angel investor and AI expert, told Huaxia Times that current technical safeguards may struggle to effectively keep up with the rapid spread of applications. “Model security does not equal system security. Security issues are complex. As OpenClaw’s application scenarios expand and user numbers grow, new security risks may continuously emerge. Developing, deploying, and updating protective measures takes time and cannot quickly cover all risks.”

He believes, “Under current technology, the ambiguity of AI agent permissions may lead to excessive data collection, and the liquidity of open-source community code increases the risk of data leaks. If maliciously exploited, AI agents could become tools for privacy theft.” He suggests industry accelerate establishing permission governance frameworks, applying the ‘least privilege’ principle to limit AI permissions.

Who benefits from the “raising lobsters” frenzy?

This wave of “lobster” enthusiasm first influenced the stock market, with many OpenClaw concept stocks rising against the trend. On March 9, UCloud’s “20cm” hit the daily limit, and QingCloud Technology rose over 15%. Dozens of listed companies in China announced they had completed adaptation or launched related services.

But the rally was short-lived. On March 11, 12, and 13, many OpenClaw concept stocks in the A-share market experienced sharp declines.

In this hype, which trends are supported by fundamentals, and which are just concept speculation?

Huo Hongyi believes that the industry trends with solid fundamentals include computing power services, enterprise application deployment, and hardware adaptation. “UCloud and others providing OpenClaw cloud deployment services, Hand Enterprise’s integration into enterprise platforms, and Loongson’s local hardware deployment—these companies leverage their technical and resource advantages and are likely to benefit from OpenClaw’s commercialization. Conversely, some companies only claim related layouts without real technical strength or business plans, and may be driven by concept hype.”

Jiang Han pointed out that the real beneficiaries are upstream hardware providers, midstream enterprise solution developers, and downstream ecosystem application developers. They provide computing resources, industry application solutions, and specific services, respectively, promoting OpenClaw’s commercialization.

Additionally, discussions like “earning 260,000 yuan in a few days by installing OpenClaw” or “on-site installation for 500 yuan per time” are spreading online. Is it really so profitable to “raise lobsters”?

“Entrepreneurs using OpenClaw to start a ‘one-person company’ have widely varying actual profits,” Huo Hongyi said. “Early on, some made profits through installation services, but as the technology became widespread, the prices for installation dropped sharply, squeezing profit margins.”

He pointed out that claims like “earning 260,000 yuan in a few days” are not sustainable or replicable. “This was mainly a result of early information asymmetry. As market competition intensifies and technical barriers lower—cloud platforms now offer one-click deployment for free, and tutorials are abundant—relying solely on installation services is no longer highly profitable. Sustainable profits now come from providing ‘deployment + operation + customization’ packages, developing proprietary skill templates, or deep industry solutions.”

Gao Zelong, vice president of the China Communications Industry Association Digital Platform Branch and expert at the China Artificial Intelligence Industry Development Alliance, told Huaxia Times that, currently, the actual profitability of entrepreneurs using OpenClaw to start a “one-person company” varies. Some entrepreneurs may have achieved high income by seizing market opportunities or offering unique products/services, but many face difficulties in profitability.

He warned that, on one hand, fierce market competition and homogeneous products/services make it hard to secure stable clients and income; on the other hand, entrepreneurs bear various costs—computing power, marketing, labor—and if income doesn’t cover these, losses occur. The market environment is constantly changing, competitors keep emerging, and consumer demand is becoming more rational. Short-term high earnings are unlikely to be sustained. Moreover, resource, capability, and experience differences among entrepreneurs mean others cannot easily replicate successful cases. When drawing lessons from successful examples, entrepreneurs should analyze their own situation rationally and avoid blind followings.

Editor: Xu Yunqian Chief Editor: Gong Peijia