A new wave of consumer AI tools has created an unexpected revenue stream: anxiety. In China, the launch of OpenClaw has been framed less as a productivity breakthrough than as a new line item in household budgets — a de facto "compute tax" that turns curiosity and fear of missing out into real spending.
The phenomenon is familiar. Parents buy myriad products and services to avoid letting their children "lose at the starting line," and adults pay for career training regardless of labour-market cycles. In the AI era those same emotions now funnel money into compute: token-based access, cloud calls and premium AI services. What used to be the preserve of trillion-dollar tech players is, for the first time, reaching ordinary consumers.
Big tech's appetite for compute set the stage. Companies from Alibaba and Tencent to Microsoft and Google have poured hundreds of billions into data centres and chips, enriching vendors such as Nvidia and Broadcom. Much of that capital expenditure lacks a clear short-term return, but executives prefer to spend rather than risk being left behind — a mindset epitomised by Larry Page's remark that he'd "rather go bankrupt than lose this race."
OpenClaw collapses that same dynamic into a consumer product. Users report that simple, everyday tasks — generating a few files, answering questions — can consume tokens that cost tens of yuan, roughly a few dollars, per interaction. Some short-form videos on platforms like Douyin show users burning thousands of yuan in a single day. For many, subscribing to human labour — for example a university student doing routine tasks — can be cheaper than running the AI constantly.
The macro data bear this out. OpenRouter reports China’s weekly large-model calls surged to 4.19 trillion tokens last week, returning the country to the top of global rankings. Behind that number are real currency flows to cloud providers and chip vendors, and micro-economies of installers, boutique consultants and course sellers who monetise users' anxiety about adoption.
Many who install OpenClaw quickly discover they do not know how to use it productively. The typical post-install script is tentative: produce emails, check calendars, read news and weather — tasks that cheaper models such as Claude or Manus handle at better cost-performance. The mismatch between expectation and utility means people "hammer at nails looking for a reason" to use the tool, racking up token bills in the process.
Commercial winners so far have not been the dreamers who hoped to build one-person empires powered by OpenClaw, but the service ecosystem around it. Technicians who install or remove the software, educators who sell short courses and influencers who package use cases capture much of the consumer surplus. The AI itself sometimes functions primarily as emotional currency: ownership confers the feeling of leadership and technical modernity rather than measurable productivity gains.
This is not a unique chapter in AI's story. Since ChatGPT appeared, an industry of training camps, wraparound membership sites and specialised tools sprung up to sell access and ease users' anxieties. The essential lesson is social and economic: people overestimate how much changes in a single year and underestimate the slow, structural shifts over a decade. OpenClaw could yet be revolutionary, but early adopters who equate possession with advantage may be paying dearly for status rather than efficiency.