A Guangzhou bank just lent against a compute bill, not a building
China spent two years building a token economy at industrial scale. This week it started building token-backed credit on top of it — while Washington is still arguing about who's allowed to referee the labs.
China's first "token loan" landed in Guangzhou, where an API call log now does the job of a property deed. Bank of China's Haizhu branch approved a three-year, 3 million yuan unsecured loan for Tengyuan Digital, a digital-marketing firm whose founder Luo Ruichang said banks previously demanded real-estate collateral his AI-servicing business does not own. The new underwriting file is a company's token bill: monthly token consumption, compute procurement contracts, API call logs — data the banks describe as naturally tamper-resistant once contracts, invoices and settlement flows are cross-checked against it. The shift follows a July notice from the People's Bank of China and eight other departments pushing credit review from "look at collateral" to "look at data," and it is spreading fast: Construction Bank Guangdong approved a 30 million yuan token-backed line for a Nansha IoT company, ICBC's Shenzhen Longgang branch priced a 5 million yuan "compute e-loan" for an embodied-AI training startup, and BOC Guangzhou now runs three token-loan products — supply, application and service — capped at 30 million yuan per borrower.
The scale underneath is why this matters beyond one province — we tracked the demand curve in August — China expects to burn 100 quadrillion tokens this year — and Beijing's E-Town district is already paying up to 50% of an enterprise's token spend under its "Token Ten Measures," with four ten-thousand-card token factories planned by 2030. What the banks concede is the catch: no mature risk model, no unified token-billing standard, no neutral third-party audit of the very logs that now carry the loan, and a live fraud incentive — idle tokens resold twice, consumption inflated to grow the credit line. The original read: Western labs compete over enterprise ARR; China is deciding what an inference bill is worth as collateral, which makes token consumption not just a cost line but a balance-sheet asset — the first financial instrument priced directly off model usage, and the first exposed to token price swings.
The ex-Anthropic researcher who set off last week's storm says Congress should let the labs referee themselves — until a real regulator exists. Jacob Coxon told NBC's "Meet the Press" on Sunday that the lab CEOs now calling for pacing are "completely genuine," and that any statute must move as fast as model releases because a national kill switch ages out mid-debate: "it'd be quite a lot of switches," he said, and a model that runs an internet-wide hacking swarm — the scenario Amodei described this month — would slip past it. The tension is his own: the man who resigned last week accusing OpenAI and Anthropic of "racing straight to self-improving superintelligence and gambling with our lives" just endorsed letting those same labs keep the whistle mid-race. We watched what happened the moment his warnings turned into a commercial liability — Silicon Valley has started punishing the AI doomers.
What to watch: whether Guangdong's token-loan book grows past the first ¥28 million of approved lines before a token-consumption data scandal gives the skeptics their case.
If your compute bill became your collateral, would you start spending tokens differently?
