The promise of autonomous wealth and digital efficiency recently swept through the Chinese technology sector, driven by a viral AI agent software known as OpenClaw. For George Zhang, a cross-border ecommerce professional based in the coastal city of Xiamen, the software represented a potential windfall despite his limited understanding of its underlying architecture. Zhang’s interest was piqued by a viral video featuring a Chinese social media influencer who demonstrated how OpenClaw could be deployed to autonomously manage stock portfolios, making complex investment decisions without human intervention. By late February, Zhang joined thousands of others in an attempt to install the software, marking the beginning of a national craze that has since evolved into a complex case study of technological hype, corporate opportunism, and the widening gap between technical and non-technical users.
The OpenClaw phenomenon has manifested in nearly every facet of Chinese urban life. In major cities, workshops dedicated to teaching the nuances of AI agent deployment have drawn crowds in the hundreds. Local governments, eager to signal their alignment with the national "AI Plus" initiative, have announced subsidies for entrepreneurs building products within the OpenClaw ecosystem. Last week, the trend reached a fever pitch when images of elderly citizens lining up at tech hubs to have the software installed on their devices went viral, highlighting the extent to which the "agentic" AI dream has permeated the public consciousness. However, as the initial excitement settles, a clearer and more sobering picture is emerging regarding the costs and capabilities of this new frontier in artificial intelligence.
The Mechanics of the "Lobster" Craze
To the Chinese public, OpenClaw is affectionately known as the "lobster," a play on the software’s name. For a typical user like Zhang, the onboarding process involved a series of financial and technical steps: renting a cloud server from Tencent and purchasing a subscription to Kimi, a prominent Chinese large language model (LLM) developed by Moonshot AI. Once configured, the agent acts as a persistent digital entity capable of performing tasks across the web.
Initially, Zhang was impressed. He watched as his "lobster" generated comprehensive market analyses based on real-time breaking news. However, the honeymoon period was short-lived. Within days, the agent’s performance degraded. Instead of detailed reports, it began producing basic outlines. When prompted to return to its original level of detail, the agent entered a loop of responding that it was "working on it" without ever delivering results.
Zhang’s experience highlights a fundamental reality of the current AI landscape: OpenClaw, while powerful, is not a "plug-and-play" solution for the layperson. "It would tell me I needed to configure the API port," Zhang noted, describing a task that requires a level of technical literacy he did not possess. Ultimately, Zhang abandoned the idea of autonomous stock trading, repurposing the tool for the simpler task of aggregating industry news to fuel a content farm on WeChat. This pivot from high-level financial strategy to basic data scraping is emblematic of the experience of many non-technical adopters.
A Chronology of the OpenClaw Boom
The trajectory of OpenClaw in China followed a classic hype cycle, accelerated by the country’s unique social media and corporate ecosystem.
- Early February 2024: Influencers on platforms like Douyin and Bilibili began showcasing OpenClaw’s ability to "self-correct" and perform multi-step tasks, such as booking travel or managing investments.
- Mid-February 2024: The "Gold Rush" phase began. Technical entrepreneurs realized they could charge for installation services. Reports surfaced of individuals earning significant sums—one technician reportedly performed over 7,000 installations at roughly $34 per instance, generating nearly $240,000 in revenue.
- Late February 2024: Major tech firms began to pivot. Seeing the massive influx of new users, companies like Tencent and Alibaba began offering free installation support at their corporate headquarters to capture the resulting cloud and API traffic.
- Early March 2024: The "Reality Check" phase. Users began to realize the staggering costs of "token burn." Unlike a standard chatbot, an autonomous agent like OpenClaw remains active, constantly pinging LLMs and consuming data units (tokens), leading to unexpected monthly bills.
The Economic Reality of Token Consumption
The primary driver behind the corporate embrace of OpenClaw is financial. For providers of cloud infrastructure and LLMs, an active OpenClaw user is far more valuable than a standard chatbot user. Poe Zhao, a tech analyst and founder of the Hello China Tech newsletter, explains the disparity: "A chatbot uses only a few hundred tokens per conversation; a single active OpenClaw instance can consume tens or even hundreds of times more tokens per day."
Because OpenClaw operates 24/7, it creates a constant stream of API calls. This explains why engineers from Tencent were seen setting up physical tables outside their offices to assist the public with free installations. Every person who successfully installs the software becomes a permanent customer of the company’s cloud services and LLM tokens.
For the average user, the costs are not insignificant. Zhang estimated his initial setup cost at approximately $30 for the server and a basic Kimi subscription. However, for complex tasks, these costs can escalate rapidly. Rain Miao, a startup founder and seasoned AI user, notes that technical proficiency is the only way to make the software cost-effective. Miao suggests a tiered approach: delegating high-reasoning tasks to expensive models like ChatGPT while offloading repetitive labor to cheaper, domestic Chinese models. Without this level of optimization, the "lobster" becomes an expensive digital pet.
Technical Barriers and User Frustration
The divide between the "tech-savvy" and the "layperson" has never been more apparent than in the OpenClaw rollout. Song Zhuoqun, a college student and intern at an AI startup, represents the frustrated majority. Despite her proximity to the industry, she found the installation process impenetrable. When she turned to Doubao—ByteDance’s AI chatbot—for help, the resulting instructions were a confusing maze of code blocks.
"I just kept asking the AI to generate a response for me, then I’d paste it over, run it, and it would run into an error," Song said. This "trial-and-error" method resulted in a non-functional installation and a sense of disillusionment. Even high-profile figures have weighed in on the frustration. Changpeng Zhao, the founder of Binance, remarked on social media that while people claim the software requires no effort after installation, most users find themselves spending all their time "tweaking that useless lobster that can’t do anything."
Corporate Fragmentation and Open-Source Friction
As OpenClaw is open-source software, the Chinese tech giants have moved to "localize" it, creating a fragmented ecosystem of proprietary versions. Today, the market includes Tencent’s QClaw, ByteDance’s ArkClaw, Moonshot’s KimiClaw, and Z.ai’s AutoClaw. While these companies claim their versions are more user-friendly and better integrated with local apps like WeChat and Alipay, critics see this as a move to "lock" users into specific corporate ecosystems.
This aggressive commercialization has drawn the ire of the original creator of OpenClaw, Peter Steinberger. He recently expressed his disapproval on social media, noting that while Chinese firms are quick to copy the features of the project, they offer little to no support to the original open-source community. This tension highlights a recurring theme in the global tech landscape: the friction between open-source innovation and corporate monetization.
Broader Implications: From Web3 to the Agentic Internet
The OpenClaw craze signals a significant shift in the Chinese digital economy. Many of the organizers now promoting OpenClaw workshops were previously involved in the cryptocurrency and Web3 space. As interest in "Metaverse" projects has waned, these communities have pivoted to what some are calling "Web4.0" or the "Agentic Internet."
The willingness of the Chinese public to pay for AI services is perhaps the most surprising outcome. Historically, Chinese consumers have been accustomed to free software in exchange for data or exposure to advertisements. The fact that thousands are now paying for cloud servers and API tokens suggests a fundamental change in consumer behavior, driven by a fear of being left behind in the AI revolution.
However, risks remain. OpenClaw has been flagged by cybersecurity experts for potential vulnerabilities, as the software requires significant permissions to operate autonomously across a user’s digital life. Despite these risks, local governments continue to offer incentives for its development, often viewing it as a symbolic marker of technological progress rather than evaluating its practical security implications.
Conclusion: The Future of the Autonomous Agent
As the dust settles on the initial OpenClaw explosion, the software is transitioning from a viral sensation to a specialized tool. For those with the technical skills to configure and optimize it, OpenClaw remains a transformative productivity multiplier. For the general public, it has served as a costly lesson in the complexities of modern AI.
The "lobster" craze may eventually be remembered as a period of collective FOMO (fear of missing out), but its impact on the Chinese tech ecosystem is undeniable. It has accelerated the adoption of paid AI services, solidified the dominance of major cloud providers, and set the stage for the next generation of autonomous digital labor. Whether these agents will eventually replace the "unpaid interns" as some internet jokes suggest, or remain a niche tool for the technologically elite, depends on how quickly the industry can bridge the gap between "viral hype" and "user-friendly reality."
About the Author
