The person who burned 9.3 million yuan in tokens in a month didn't find an answer

Token burning becomes a workplace spectacle

It has been reported that an internal screenshot circulating online showed a single account consuming about 603 billion AI tokens in 30 days — a bill reportedly totalling roughly ¥9.3 million (about US$1.3 million). Tokens have leapt from a technical unit into corporate KPI, consumer telecom bundles and even campus "donations." China's three state carriers — China Mobile (中国移动), China Unicom (中国联通) and China Telecom (中国电信) — are now selling token-laced plans. But what does a token actually measure?

Perverse incentives at Big Tech

Western firms are not immune. The Financial Times reported that Amazon (亚马逊) tracked employee token use and that this became a de facto metric, prompting workers to "token-max" — spinning up armies of trivial AI agents to inflate consumption. Meta reportedly saw an internal dashboard ranking top token consumers before it was taken down, and anonymous employees on workplace forums described using agents to roast product managers or plan vacations simply to climb leaderboards. GitHub’s former CEO Nat Friedman and others have also shared anecdotes of AI agents burning compute on banal tasks. These accounts indicate that metrics meant to accelerate AI adoption can morph into gamified competitions that reward waste.

A symptom of bigger tensions

This is not just about comedy screenshots. Token-led incentives collide with real costs and policy pressures: export controls on high-end GPUs and global scrutiny of generative AI make inefficient use of scarce compute a strategic and regulatory concern. In China, even charitable storytelling has shifted — three alumni reportedly "donated" 20 billion tokens to Zhengzhou Sias University (郑州西亚斯学院), a gesture mocked online using a platform price benchmark. Observers warn that treating tokens as a vanity KPI risks creating new, meaningless busywork while obscuring whether AI actually delivers value.

So what should companies measure?

Companies chasing token counts may feel like they are riding the AI wave, but burning through tokens does not guarantee better products or less toil. If the point was to reduce "bullshit work," why is AI generating more of it? The spectacle of millions of yuan in token bills should prompt a simple question: are firms measuring meaningful impact — or just measuring the fire?