Why this story

Why it matters
Businesses buying model access need to know what they paid for. Comparable categories would let a customer challenge a bill or distinguish a cheaper service from one that simply counts different items.
Attention gap
The publisher checked China’s data agency, the standards registry and Bloomberg’s report on the speech. A Reuters report about retail demand predates the announcement and supplies market context.

At a computing conference in Changsha on November 20, National Data Administration director Liu Liehong described businesses that bundle access to computing power, models and data. Customers can buy the resulting service by the call, with usage measured in tokens. Liu said the agency would strengthen billing, measurement and quality standards, and warned against duplicating low-quality infrastructure. The agency posted his remarks the following day.

A proposed national counting standard shows why a common bill is difficult. A model can consume resources processing a prompt, reusing a cached result or producing an answer. The project covers input, output, reasoning and multimodal tokens and proposes a uniform reporting format. The standards registry says users could compare counts from different providers, while auditors and regulators could check usage and costs.

The registry lists a 15-month project and an October 30 end to its public notice period. Drafters include Zhejiang Lab, the China Academy of Information and Communications Technology, Alibaba Cloud, Baidu and DeepSeek. The agency’s speech set no price or binding implementation date for providers.

The consumer market gives the proposal immediate relevance. Reuters reported that Tmall’s AI-token sales rose more than 700% from October 15 through November 5 compared with September 15 through October 5. A livestreamer sold 100,000 monthly QwenWork subscriptions in 30 minutes. These are figures from a shopping campaign, not a national measure of AI usage.

What the evidence cannot tell us

  • The agency announced a standards plan. The proposed counting document is a recommended-standard project; neither source establishes a mandatory price, enforcement date or rule for foreign suppliers.

What to watch

Publication of the counting draft and any billing instrument that sets obligations, effective dates or enforcement for providers.

Read the sources. Credit the work.

The reporting and documents behind this brief, linked directly to the source. Verified publisher and author credits appear where available; other entries show their actual website address. Several links may draw on the same underlying reporting or research.

  1. National Data Administration of China

    Liu Liehong addresses the 2026 World Computing Conference (opens in a new tab)

    nda.gov.cn · November 21, 2026

    Primary document / sourceContent read for this publication · November 22, 2026

    Publisher credit checked November 22, 2026. View credit on the source page (opens in a new tab).

  2. China’s National Public Service Platform for Standards Information

    Intelligent computing — Token service — Part 2: Counting rules (opens in a new tab)

    std.samr.gov.cn · September 30, 2026

    Primary document / sourceContent read for this publication · November 22, 2026

    Publisher credit checked November 22, 2026. View credit on the source page (opens in a new tab).

  3. Bloomberg via Moneycontrol

    China to tighten AI token measurement, billing standards as token factories emerge (opens in a new tab)

    moneycontrol.com · November 21, 2026

    Credited original publisher: Bloomberg (opens in a new tab)

    Reporting / analysisContent read for this publication · November 22, 2026

    Publisher credit checked November 22, 2026. View credit on the source page (opens in a new tab).

  4. Reuters

    China’s Singles Day shopping festival puts AI selling power to the test (opens in a new tab)

    reuters.com · November 11, 2026

    Reporting / analysisContent read for this publication · November 22, 2026

    Publisher credit checked November 22, 2026. View credit on the source page (opens in a new tab).

This story was written and checked using AI against the linked evidence. No routine human review is claimed. Read the method.

Suggest an update or correction
Published corrections