AI Agents Northeastern and Stanford Give Agents a Git History: Shepherd Lifts Pair-Coding Pass Rates From 28.8 to 54.7 Percent Coding Assistants China Telecom Trains a 285B-Parameter Agentic Model on 8,192 Ascend 910C Chips With No Nvidia Silicon Anywhere in the Stack News Google Cuts 2027 Broadcom-Designed TPU 8i Orders by About a Million Units and Pushes Volume Toward MediaTek Business China's Data Regulator Will Set National Rules for Metering and Billing AI Tokens After Daily Use Tripled to 500 Trillion Business Shanghai's Yuanshi Technology Raises Nearly RMB 300 Million for AI Virtual Cells, Two Weeks After the Last Chinese Record Fell Security GAO Finds No Widely Accepted Test for Generative AI Cyber Skills and Hands Congress 16 Options, Including Human Sign-Off on Consequential Actions News Microsoft Turns Three USB4-Cabled PCs Into One AI Machine: Qwen3-235B Runs at Up to 22 Tokens a Second Across 384GB of Ordinary Memory Business Dell, HPE, Lenovo and Supermicro Are Demanding Up to 45% More for Signed AI Server Deals, and Trading Firms Are Paying AI Agents Northeastern and Stanford Give Agents a Git History: Shepherd Lifts Pair-Coding Pass Rates From 28.8 to 54.7 Percent Coding Assistants China Telecom Trains a 285B-Parameter Agentic Model on 8,192 Ascend 910C Chips With No Nvidia Silicon Anywhere in the Stack News Google Cuts 2027 Broadcom-Designed TPU 8i Orders by About a Million Units and Pushes Volume Toward MediaTek Business China's Data Regulator Will Set National Rules for Metering and Billing AI Tokens After Daily Use Tripled to 500 Trillion Business Shanghai's Yuanshi Technology Raises Nearly RMB 300 Million for AI Virtual Cells, Two Weeks After the Last Chinese Record Fell Security GAO Finds No Widely Accepted Test for Generative AI Cyber Skills and Hands Congress 16 Options, Including Human Sign-Off on Consequential Actions News Microsoft Turns Three USB4-Cabled PCs Into One AI Machine: Qwen3-235B Runs at Up to 22 Tokens a Second Across 384GB of Ordinary Memory Business Dell, HPE, Lenovo and Supermicro Are Demanding Up to 45% More for Signed AI Server Deals, and Trading Firms Are Paying

China's Data Regulator Will Set National Rules for Metering and Billing AI Tokens After Daily Use Tripled to 500 Trillion

China's Data Regulator Will Set National Rules for Metering and Billing AI Tokens After Daily Use Tripled to 500 Trillion

China's National Data Administration plans national standards for how AI tokens are metered, billed and quality-tested, director Liu Liehong said at the World Computing Conference in Changsha on Nov 20. Daily token consumption in China rose from more than 140 trillion in March to more than 500 trillion by June, and so-called token factories now resell rented compute as metered AI service.

China will write national standards for how artificial intelligence tokens are measured, billed and quality-tested, the country's data regulator said, moving to formalise the unit that has quietly become the pricing basis for most of the country's AI services. Liu Liehong, director of the National Data Administration, announced the plan in an opening address at the 2026 World Computing Conference in Changsha, Hunan province, on Nov 20.

The numbers behind the decision are large. Daily token consumption in China rose from more than 140 trillion in March 2026 to more than 500 trillion by June, according to figures cited by the agency and reported by Bloomberg. In two years the country's daily volume has grown roughly a thousandfold from the roughly 100 billion calls recorded at the start of 2024.

Liu framed tokens as the connective tissue of the AI economy rather than a technical metric. Tokens, he said, link compute, models and data into a single value chain, making them the anchor for pricing intelligent services and the most visible expression of what data is worth once it has passed through a model. The term itself was only recently settled: in March the China National Committee for Terms in Sciences and Technologies recommended "ci yuan" as the standard Chinese rendering of token.

A new class of supplier has grown up around that unit. Liu described the rapid rise of "token factories", businesses that bundle rented computing power, model access and proprietary data sets and resell the output as metered tokens, including through banks and telecom carriers. That model, he said, converts what used to be infrastructure rental into a delivery format that can be counted, invoiced and audited per call.

The regulator's stated next steps are to widen the application scenarios for token services and to set standards for metering, billing and quality evaluation. Liu also warned against "low-quality duplicate construction", language that in Chinese industrial policy usually signals an intent to thin out marginal players rather than to expand capacity. The agency did not say when the standards would be published, and no draft has been circulated publicly.

The timing is not accidental. Chinese model developers have spent the past quarter competing almost entirely on price per token, and the country's largest labs have signed long-term compute contracts whose economics depend on assumptions about how much inference demand will materialise. A common measurement standard would make those comparisons legible to enterprise buyers, but it would also make discounting harder to disguise.

For foreign observers, the more consequential detail is structural. Beijing is treating token consumption as an economic indicator to be standardised and steered, in the same way it once standardised electricity tariffs and mobile data pricing. That gives the state a measuring instrument for AI adoption at a moment when the private market's own numbers are inconsistent and largely self-reported.

What this means: a national metering standard is the unglamorous prerequisite for AI becoming a utility, and China is writing it before most Western regulators have even settled on a definition of the unit. The risk is that a single mandated metric, set while the technology is still changing shape, ends up measuring the wrong thing.

Comments (0)

Log in to join the discussion

Log In

No comments yet