China’s OpenClaw Boom Is a Gold Rush for AI Companies

George Zhang thought OpenClaw could make him rich, even though he didn’t really understand how the viral AI agent software worked. But he saw a video of a Chinese social media influencer demonstrating how it could be deployed to manage stock portfolios and make investment decisions autonomously. Zhang, who works in cross-border ecommerce in the Chinese city of Xiamen, was intrigued enough that he decided to try installing OpenClaw in late February.

Zhang is one of the many people in China who got swept up in the craze over OpenClaw recently. Workshops teaching people how to use the AI agent have popped up in cities across the country, drawing crowds of hundreds. Tech companies are racing to integrate OpenClaw into their platforms, while local governments have announced subsidies for entrepreneurs building products with it. Late last week, images of grandpas and grandmas lining up to install the software went viral across the internet.

After renting a cloud server from Tencent and buying a subscription to the Chinese large language model Kimi, Zhang could start chatting with his OpenClaw agent, or his “lobster,” as many Chinese people call theirs. At first, Zhang tells me, he was impressed by the AI agent as he watched it quickly generate a long market analysis based on the latest breaking news. But a few days in, his lobster started slacking off, and it would generate only a basic outline of market trends instead of a detailed report. He asked OpenClaw to generate something like what it had done on the first day, to which the agent perpetually responded that it was “working on it” before never returning any results.

Zhang’s conclusion was that OpenClaw is not designed for people like him who don’t have any coding skills. “It would tell me I needed to configure the API port. But that’s a technical task, not something I can do unless I had a tutorial walking me through it step-by-step,“ he says. In the end, he gave up on letting his lobster trade stocks, settling instead on asking it to aggregate AI industry news, which he used to build a social media content farm on WeChat.

This week, I checked in with half a dozen users of OpenClaw in China about their experiences with the agent, and a clear picture of division emerged between the adopters who are technologically savvy and those who are not. People who are proficient in AI see OpenClaw as a game changer in productivity, but those with no technical background feel they were promised a miraculously powerful AI product that ultimately didn’t deliver. But by the time the bubble burst, they had already started paying for cloud servers and LLM tokens.

The real driver of the OpenClaw mania in China isn’t everyday users, but rather the Chinese companies that stand to benefit financially from its widespread adoption. Major tech firms like Tencent, Alibaba, ByteDance, Minimax, Moonshot, and Z.ai all saw the AI productivity FOMO as a rare chance to get normal people to start paying for AI services, and they are reaping the biggest rewards from it.

“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,” says Poe Zhao, a tech analyst and founder of the newsletter Hello China Tech. Every new user of OpenClaw is someone who’s paying 24/7 for LLM API calls. “That’s why Tencent engineers were setting up tables outside headquarters to help people install the software for free,” he says.

“I Couldn’t Understand Any of It”

Song Zhuoqun, a college student in China, says she started running into problems with OpenClaw as soon as she tried installing it. Song is a social media intern at an AI startup but has no programming experience, so figuring out how to get OpenClaw running turned out to be difficult. She asked Doubao, ByteDance’s popular AI chatbot, to generate a step-by-step tutorial for her, but it wasn’t much help.

“There were pages full of code, and I couldn’t understand any of it. 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,” she says. The installation ended up being the most frustrating part of trying out OpenClaw for Song, and she didn’t feel like she learned anything from it.