OpenClaw's Second Hundred Days Just Started
1,000 people. 17 cities. 40 days. OpenClaw's next phase.
One hundred days ago, OpenClaw was a GitHub project. On Saturday in Shenzhen, a thousand people packed into a robotics venue for an event co-hosted by the Longgang district government and Moonshot AI’s Kimi. Telecom carriers set up service points outside. Installation booths ran out of appointment slots after releasing 300 numbers. Volunteers turned away anyone who showed up without a laptop. One staffer told the crowd: “Uninstall your antivirus software first, then get back in line.”
The same day, Tencent Cloud announced a 40-day nationwide deployment campaign spanning 17 cities. Technical teams will offer free OpenClaw installation, model configuration, and security setup across Shenzhen, Shanghai, Beijing, Guangzhou, Hangzhou, Chengdu, Wuhan, and beyond. The product lineup covers Tencent’s full agent stack: its Lighthouse cloud hosting, QClaw desktop app, WorkBuddy enterprise agent, and supporting cloud infrastructure. Five days earlier, on March 9, Tencent had launched all three agent products in a single day, sending its stock up 7 percent and pushing its market cap back above HK$5 trillion.
These events share a common signal. The spontaneous phase of OpenClaw adoption in China is over. What follows is organized, strategic, and accelerating.
Three Layers, One Race
I have spent this week analyzing why OpenClaw’s reception in China has been so structurally different from a typical open-source hype cycle. The answer runs deeper than enthusiasm. Every major layer of China’s AI ecosystem found specific, self-interested reasons to support this framework. Saturday’s events are the direct result.
The cloud economics layer. China’s major tech companies committed an estimated $60 billion in combined AI infrastructure spending over the past year. Standard chatbot usage was never going to generate enough inference demand to justify that investment. A single active OpenClaw agent consumes tens to hundreds of times more tokens per day than a chat session. Every installed instance becomes a round-the-clock revenue source for cloud providers. Tencent deploying technical teams across 17 cities is a customer acquisition campaign dressed as community service. I mapped the full supply-side logic, including why the cost advantage of Chinese open-source models creates a self-reinforcing consumption loop, in OpenClaw Conquered China in 100 Days.
The platform strategy layer. Among all the companies embracing OpenClaw, Tencent moved with unusual speed and breadth. The reason traces back to a problem the company spent over a year failing to solve. Its Hunyuan model ranked 68th on major benchmarks. Its consumer AI app trailed ByteDance and Alibaba by wide margins. OpenClaw shifted the competitive axis from model quality to interface control. And Tencent controls the messaging platform where 1.4 billion people spend their days. In Why Tencent Needs OpenClaw More Than Any Other Chinese Tech Giant, I analyzed how Tencent’s three-product launch on Monday maps onto a deliberate, layered strategy to bring AI agents into WeChat without breaking WeChat. That strategy includes a confidential internal project that could reshape how the app functions by Q3 2026.
The policy layer. Saturday’s Longgang event was not just a tech meetup. It was a district government co-production. Over the past two weeks, at least 6 districts and development zones across China rolled out formal support measures for OpenClaw adoption. Subsidies reached up to Rmb20 million ($2.8 million) per entity per year in some jurisdictions. The shared policy target: OPC, or One Person Company, a new category of economic actor built around individual developers using AI agents as their primary workforce. In OpenClaw and China’s Local Government Experiment, I examined which cities are building genuine ecosystems versus planting flags, and why the compute economics underlying these subsidies remain dangerously unresolved. The detail that one Shenzhen programmer burned through Rmb12,000 ($1,660) in 3 days from a misconfigured agent is not an outlier. It is a structural feature of how agent workloads consume resources.
What Happens Next
The pattern across all three layers points in one direction. Cloud vendors, LLM startups, local governments, and platform companies are all racing to convert OpenClaw’s viral moment into durable infrastructure before user enthusiasm fades. The first wave of installation-service customers is already uninstalling. The companies and cities that locked in sustainable economic relationships during this window will carry the advantage forward. Those that merely rode the attention cycle will not.
This email gave you the pattern. Understanding the mechanisms underneath requires going deeper. This week I published three Premium analyses covering the full stack: the supply-side economics driving cloud vendors and LLM startups, Tencent’s specific strategic calculus including a confidential WeChat project, and the local government policy race where compute subsidies may be creating more problems than they solve. Together, they map the forces shaping who wins and who wastes money in China’s AI agent era.