On March 6, nearly a thousand people queued outside Tencent's headquarters in Shenzhen to have engineers install OpenClaw, an open-source AI agent, onto their devices for free. Appointment slots vanished within an hour as crowds brought everything from MacBooks to mini PCs, eager to deploy the autonomous agent that connects to large language models through API calls.

OpenClaw operates across messaging platforms including Slack, WhatsApp, Telegram, and Feishu, executing multi-step tasks like web browsing, code debugging, and calendar management. Built by Austrian engineer Peter Steinberger, the project reached GitHub's most-starred repository status in roughly 100 days—surpassing Linux's 30-year milestone. Of 142,000 tracked agents, nearly half originated from China, where paid installation services emerged overnight at $7 to $100 per setup.

The structural story lies not in user enthusiasm but in supply-side incentives. Chinese tech giants spent an estimated $60 billion on AI infrastructure, creating pressure to find sustained inference demand. OpenClaw resets consumption math: a single configured instance burns through tens to hundreds of times more tokens daily than standard chatbot usage. Every installation becomes a round-the-clock source of API calls, flowing directly into cloud vendor revenue and creating a self-reinforcing loop where cheaper models drive more agent usage.