George Zhang, a cross-border e-commerce professional in Xiamen, China, found himself swept up in the recent fervor surrounding OpenClaw, a viral AI agent software. While he admitted to not fully grasping its intricate workings, a compelling demonstration by a Chinese social media influencer—showcasing OpenClaw’s potential for autonomous stock portfolio management and investment decisions—piqued his interest. By late February, Zhang, like countless others across China, decided to install the enigmatic AI agent, hoping for a lucrative outcome.
This widespread enthusiasm for OpenClaw manifested in numerous ways throughout China. Workshops dedicated to teaching its deployment quickly sprang up in major cities, attracting hundreds of attendees eager to harness its perceived power. Tech companies scrambled to integrate OpenClaw into their existing platforms, recognizing its potential to drive user engagement and service adoption. Local governments, keen to foster innovation and attract entrepreneurial talent, announced subsidies for those developing products powered by the AI agent. The phenomenon reached a cultural apex when images of elderly citizens, affectionately dubbed "grandpas and grandmas," lining up to install the software went viral, underscoring the broad appeal and accessibility that OpenClaw was marketed to possess.
To embark on his OpenClaw journey, Zhang rented a cloud server from Tencent, a leading Chinese technology conglomerate, and secured a subscription to Kimi, a prominent Chinese large language model. This setup allowed him to interact with his OpenClaw agent, which many Chinese users affectionately referred to as their "Lobster." Initially, Zhang was impressed. His agent swiftly generated a comprehensive market analysis, drawing upon the latest breaking news. However, this initial success proved ephemeral. Within a few days, the agent’s performance declined, regressing to producing only basic outlines of market trends instead of the detailed reports he had come to expect. When Zhang requested a return to its initial, more robust output, the agent repeatedly responded with a generic "working on it" before failing to deliver any results.
Zhang’s experience led him to conclude that OpenClaw was not designed for individuals without coding proficiency. He recounted encountering technical hurdles, such as requests to "configure the API port," tasks that were beyond his skill set without explicit, step-by-step guidance. Ultimately, he abandoned the idea of using his "Lobster" for stock trading. Instead, he repurposed it to aggregate AI industry news, which he then utilized to build a social media content farm on WeChat.
A survey of half a dozen OpenClaw users in China revealed a distinct bifurcation in experiences, largely correlating with their technological aptitude. Those proficient in AI perceived OpenClaw as a significant productivity enhancement, a genuine game-changer. Conversely, users with no technical background often felt they had been promised a revolutionary AI product that failed to deliver on its ambitious claims. By the time the initial hype began to dissipate, these users had already incurred costs for cloud servers and large language model tokens.
The true engine behind the OpenClaw mania in China was not individual users seeking personal enrichment, but rather the nation’s major technology companies. Firms such as Tencent, Alibaba, ByteDance, Minimax, Moonshot, and Z.ai recognized the "fear of missing out" (FOMO) surrounding AI productivity as a unique opportunity to encourage mainstream consumers to begin paying for AI-related services. These companies stood to gain the most from this widespread adoption.
Poe Zhao, a technology analyst and founder of the newsletter Hello China Tech, explained the economic underpinnings of this phenomenon. "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," Zhao stated. Each new OpenClaw user represented a potential 24/7 revenue stream for LLM API calls. This financial incentive, Zhao noted, likely explained why Tencent engineers were seen setting up tables outside their headquarters, offering free assistance with software installation.
"I Couldn’t Understand Any of It": The Installation Hurdle
Song Zhuoqun, a college student in China and a social media intern at an AI startup, encountered immediate difficulties upon attempting to install OpenClaw. Despite her exposure to the AI industry, she lacked programming experience, making the setup process a significant challenge. She sought assistance from Doubao, ByteDance’s popular AI chatbot, requesting a step-by-step tutorial. However, the generated guide proved largely unhelpful.
"There were pages full of code, and I couldn’t understand any of it," Song recounted. "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, so I’d try a new response." For Song, the installation phase was the most frustrating aspect of her OpenClaw experience, leaving her with no discernible new knowledge.
Song’s predicament mirrored that of many who succumbed to the OpenClaw FOMO. In China, the software was marketed as an accessible tool for ordinary individuals to benefit from AI advancements. However, the reality proved far more complex. Even Changpeng Zhao, the multibillionaire founder of the cryptocurrency exchange Binance, expressed his disillusionment on social media. He lamented that while users were promised a seamless experience post-installation, their time was predominantly consumed by "tweaking that useless lobster that can’t do anything."
Rain Miao, a Chinese startup founder who actively utilizes AI agents for software development, offered a stark assessment for non-technical users. "If you still can’t figure out how to install it after a long time, and you don’t even know how to handle the basic permissions, then you’re probably better off not installing it at all," he advised. Miao suggested that less technically inclined users would find greater benefit in tools like Claude Cowork, which, despite offering comparable functionality, garnered significantly less attention in China.
The Hidden Costs: Token Consumption and Cloud Dependencies
The majority of non-technical OpenClaw users found their personal computers ill-equipped for the software’s demands. Their machines lacked the necessary compatibility with OpenClaw’s working environment and the computational power to run AI models locally. Consequently, they were compelled to rent cloud servers and subscribe to cloud-based LLM models to operate the agent. While some users in Silicon Valley opted for more powerful local hardware like Mac Minis, this solution represented a considerably higher upfront investment.
Zhang provided a breakdown of his OpenClaw expenses. Following recommendations from online tutorials, he initially rented a cloud server from Tencent for a year. He then paid for a monthly subscription to Kimi to access its API and a specific allocation of tokens. The initial setup cost him approximately $30. This figure would have been substantially higher had he utilized OpenClaw for more complex tasks that consume a greater volume of tokens.
Miao indicated that running OpenClaw at a reduced cost is feasible, but it necessitates software programming expertise to identify and implement workarounds. He recommended a strategic approach: delegating the most challenging tasks to more capable, albeit more expensive, LLMs like ChatGPT, while assigning repetitive operations to more economical domestic Chinese AI models. Owning a powerful computer capable of local processing further contributed to his cost savings.
In recent weeks, a humorous observation has gained traction on Chinese social media: OpenClaw might soon be superseded by unpaid interns. The analogy suggests that while AI agents incur tangible costs in the form of tokens, the allure of internship opportunities can incentivize free labor from students, effectively achieving a similar outcome at no direct monetary expense.
The Architects of the Frenzy: Identifying the Real Winners
The OpenClaw phenomenon offers a crucial insight into the Chinese market: ordinary citizens are demonstrably willing to pay for AI services. This willingness is particularly noteworthy given the prevailing consumer behavior in China, where software is often obtained freely in exchange for personal data and exposure to advertising. The widespread adoption of OpenClaw, however, revealed an enthusiastic, or perhaps desperate, consumer base willing to pay for cloud infrastructure and API access without hesitation. This momentum has been eagerly embraced by Chinese technology companies, which have actively promoted free installation services and live-streamed tutorials to capitalize on the trend.
Furthermore, as OpenClaw is an open-source project, nearly every major Chinese tech firm has moved swiftly to develop and release their own customized versions. Examples include Tencent’s QClaw, ByteDance’s ArkClaw, Moonshot’s KimiClaw, and Z.ai’s AutoClaw. While these companies assert that their versions offer enhanced ease of installation and seamless integration with popular applications, their underlying strategy appears to be the entrenchment of users within their respective product ecosystems.
Peter Steinberger, the original creator of OpenClaw, publicly expressed his disapproval of these proprietary adaptations. "They copy yet they don’t support the project in any way," he commented on social media platform X, in response to news of Tencent localizing OpenClaw features, accompanying his statement with a somber emoji.
A Lucrative Ecosystem: Who Else is Cashing In?
A significant overlap exists between the organizers of OpenClaw promotional events in China and those who orchestrate cryptocurrency gatherings. These communities, previously focused on Web3.0, appear to be pivoting their focus towards the concept of "Web4.0," or the agentic internet.
Savvy technical engineers have also identified a lucrative niche market by offering OpenClaw installation services to others. One individual reported to MIT Technology Review that he had successfully installed the AI agent over 7,000 times, charging approximately $34 per installation.
Even local Chinese governments have demonstrated agility in capitalizing on the OpenClaw hype. Despite the software’s widely acknowledged security risks, at least five local governments in China have proactively allocated funds to OpenClaw developers. This move, while appearing as an official endorsement, is more accurately interpreted as an opportunistic strategy by these governments to signal their receptiveness to technological innovation and their capacity to engage with the tech industry’s language. This mirrors a trend observed in 2022, when local governments promoted the Metaverse by subsidizing companies to create digital replicas of their cities, a venture whose long-term tourism impact remains questionable.
The OpenClaw frenzy in China underscores a significant shift in consumer behavior and a robust business model for AI services. While the initial promise of effortless wealth creation may have eluded many, the underlying infrastructure and the companies providing these services have emerged as the true beneficiaries of this AI agent gold rush. The long-term implications suggest a deepening integration of AI-powered agents into daily life, albeit with a clearer understanding of the technical prerequisites and the economic landscape that underpins this rapidly evolving sector.