Are you buying an Apple Mac mini just to "farm shrimp"?

Yangcheng Evening News Reporter Pan Liang

Recently, the open-source AI agent OpenClaw (nicknamed “Lobster” for its red lobster icon) has become a global sensation. This craze, dubbed by netizens as “shrimp farming,” unexpectedly ignited the hardware market. Among them, the Apple Mac mini, with its excellent performance and highly competitive price, has become the preferred device for localized deployment of this AI agent, leading to a worldwide rush to buy.

Currently, the shipping time for Mac mini on Apple’s official website has extended to over two weeks, e-commerce platforms are generally out of stock, and secondary market prices are soaring. The surge driven by AI software reflects a deeper industry trend: AI evolving from a “tool” to a “partner.”

“Physical isolation is the real necessity”

In this “Lobster fever,” consumers choose Mac mini not only for cost-effectiveness but also due to concerns over data security. Angel investor and senior AI expert Guo Tao pointed out that deploying OpenClaw on main devices poses significant risks for ordinary users. Many opt to purchase a dedicated Mac mini for deployment to achieve physical risk isolation.

Data shows that Mac mini M4 sales have surged by 300%, with second-hand prices jumping from 2,800 yuan to over 4,000 yuan, and the Pro version commanding a premium of up to 1,900 yuan. A tech-savvy user, Mr. Chen, who successfully bought a stock unit, said, “To keep data off the cloud, I’m willing to pay a premium for this ‘dedicated brain.’ Since OpenClaw requires high-level system permissions, deploying it on main devices is risky. A dedicated Mac mini can meet performance needs while balancing security and efficiency through physical isolation.”

Guo Tao believes this is not only a victory for Mac mini sales but also a strategic move by Apple’s M-series chips in the era of localized AI. It proves that lightweight, high-efficiency desktop hardware is sufficient to meet ordinary users’ demands for local AGI (Artificial General Intelligence).

Zhang Chi, Vice President of Technology at VML China, shared his experience. He specifically installed “Lobster” on a separate computer rather than his daily work machine because OpenClaw requires access to core privacy permissions like email and contacts. Using a dedicated device is safer.

This “dedicated machine” demand aligns perfectly with the characteristics of the Mac mini M4. It features the high-smoothness, low-power M4 chip, and importantly, Apple has not increased its price. Additionally, macOS offers better compatibility with OpenClaw, making it an ideal sandbox environment for “shrimp farming.” Data from the Poizon app shows that many models of Mac mini M4 are sold out, with prices increasing by up to 649 yuan in the past week.

High risk of information leakage

The popularity of OpenClaw signifies more than just a hardware trend; it marks a fundamental shift in AI paradigms. Zhong Junhao, Secretary-General of the Shanghai Artificial Intelligence Industry Association, believes that OpenClaw’s pursuit of strong operability and end-to-end results indicates AI is moving from an assistant providing “recommendations” to a work partner that “executes” tasks. Tian Feng, Director of the Fast and Slow Thinking Institute, pointed out that such intelligent agents are evolving from “Copilot” to “Agentic AI,” which will activate enormous digital economic value, but risks and benefits will coexist long-term.

However, behind this tempting “Lobster,” dangers have already emerged. The Cybersecurity Threats and Vulnerabilities Information Sharing Platform of the Ministry of Industry and Information Technology issued an emergency warning, stating that OpenClaw has a very high risk of information leakage under default settings. Over 270,000 instances of OpenClaw are exposed on the internet, with many vulnerable to unauthorized access.

Zhou Hua, CTO of a UK-based software company, warned that the agent can directly invoke system permissions. If not properly managed, the convenience could significantly increase risks, potentially leading to data leaks or permanent loss. Previously, Meta’s AI security director experienced a dramatic incident: she authorized OpenClaw to analyze her emails, but the AI forcibly deleted over 200 emails, ignoring her multiple commands to stop, and she had to cut power to stop it.

Liu Xingliang, member of the Information and Communication Economy Expert Committee of the Ministry of Industry and Information Technology, reminded that the more AI resembles an assistant, the greater the potential risks. Security boundaries cannot be ignored. For ordinary users, before “Lobster” matures fully and is regulated with compliance and safety measures, maintaining a calm mindset may be more important than blindly following the trend of “shrimp farming.”