China’s National Data Administration will improve standards for metering, billing and quality testing of AI tokens, its director Liu Liehong said on 20 November. A token is a small chunk of text, sound or image that an AI model reads or writes, and it is how AI use gets measured and charged.
Daily token use in China passed 500 trillion by June, per agency data cited by Xinhua, up from more than 140 trillion in March, as Liu said then according to Yicai via The Next Web. That is about 3.6 times in three months, by our arithmetic. Bloomberg first reported the standards plan.
Business Pill · Business Engineer
An evergreen Business Engineer explainer. It teaches a general idea only and does not discuss this report.
The key insight: China’s data regulator plans standards for metering, billing and testing AI tokens as daily use topped 500 trillion by June, about 3.6 times March’s level, and Guangdong has signed a 12-department plan with numbered targets (agency data; our arithmetic).
What Liu Said
Liu spoke at the opening of the 2026 World Computing Conference in Changsha, hosted by the Hunan provincial government. In the agency’s account, he called tokens the “value anchor” of the digital economy, linking computing power, models and data.
He described “token factories” that turn rented computing power, models and data into token production services. Xinhua has compared a token factory to “the AI era’s equivalent of a power plant”: customers pay for the tokens they use rather than renting chips or running models themselves.
Next, the agency said, it will widen uses for token services, improve standards for token metering, billing and quality evaluation, and “resolutely avoid low-quality duplicate construction”. It gave no date for the standards, The Next Web noted.

Guangdong Moves First
On 21 November, when it posted its account of the speech, the agency also republished a two-year token-economy plan that 12 Guangdong provincial departments signed on 16 November, including the local central-bank branch and the banking regulator.
It sets minimum targets: at least 5 benchmark token factories plus 6 more of two specialised kinds, 100 showcase uses, 50 uses that sell tokens abroad, 5,000 industry datasets and a database of 2,000 token-focused firms.

It also plans a provincial token trading platform, token standards covering compliance, trading, metering and billing and performance testing, and certificates for token transactions that banks could use when lending. State-owned firms are to explore buying token services together to cut costs.
Why It Matters
As we read it, a common unit would let officials price and track AI use the way they track electricity. Rhodium Group projects China’s big AI firms will spend $139 billion on AI infrastructure this year, while all its AI models earn about $10.7 billion a year, per the same report.
Some demand also crosses borders. Chainalysis found China-linked payments to proxy services for US AI models spike after US model launches.
The Structural Read
As we read it, Beijing is treating tokens like electricity: a unit it can price, track and plan around.
As we read it, the token-transaction certificates in Guangdong’s plan are meant to make AI usage something banks can lend against.
National Data Administration, on token standards
“resolutely avoid low-quality duplicate construction”
Three Implications
FOR AI MODEL SELLERS IN CHINA Billing and quality standards for tokens are coming, with no date yet (National Data Administration).
FOR LENDERS Guangdong plans token-transaction certificates that banks could use when lending (Guangdong plan).
FOR THE MARKET Daily token use topped 500 trillion by June, about 3.6 times March (agency data; our arithmetic).
What Is Not Established
The national standards have not been published and no date was given. The 140 trillion and 500 trillion figures are “more than” floors from different months, so the 3.6 times is approximate. Guangdong’s numbers are targets, not results. The translations from Chinese are ours.
The Bottom Line
China’s data regulator says it will set rules for measuring, billing and testing AI tokens, as daily use topped 500 trillion by June, about 3.6 times March’s level (agency data; our arithmetic), and Guangdong has already signed a 12-department plan with targets. The full Weekly Map covers how geopolitics reshapes the AI build: This Week in the AI Supercycle.
97,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. Liu’s remarks and the Guangdong plan are read in full in Chinese on the National Data Administration’s website; translations are ours. The 500 trillion figure is agency data as cited by Xinhua; the 140 trillion figure is Liu’s March remark as reported by Yicai and relayed by The Next Web. Bloomberg first reported the plan; we read its headline only. The 3.6 times is our arithmetic.
Sources: National Data Administration, Liu Liehong at the 2026 World Computing Conference, posted 21 Nov 2026 (Chinese) · Guangdong Government Service and Data Administration and 11 departments, Token Economy Action Plan 2026-2027, 16 Nov 2026 (Chinese) · The Next Web, China’s data regulator plans standards for measuring and billing AI tokens, 21 Nov 2026 · Xinhua via People’s Daily Online, China rolls out token factories, 13 Oct 2026 · The Business Engineer, This Week in the AI Supercycle, 21 Nov 2026








