For years, a company's most valuable asset sat off the books. On January 1, 2024, China changed that — and the first beneficiary may be the people training AI.
The rule that made data an asset
On August 21, 2023, the Ministry of Finance (财政部) released the Interim Provisions on Accounting Treatment for Enterprise Data Resources (企业数据资源相关会计处理暂行规定). They took effect on January 1, 2024. The core move is deceptively simple: under defined conditions, data a company controls can be recorded as an intangible asset or inventory on the balance sheet, with a dedicated "data resources (数据资源)" line sitting under intangible assets, inventory, and development expenditure.
Data stopped being a cost you expensed and forgot. It became something you carry, value, and defend.
How "入表" works
The Chinese shorthand is 入表 (rù biǎo) — "onto the books." In practice:
- Data a firm uses to deliver services, meeting the criteria, is recognized as an intangible asset and amortized as it is consumed.
- Data a firm holds to sell, meeting the criteria, is recognized as inventory and counted when sold.
- Firms are invited to voluntarily disclose what they will not formally book: application scenarios, the type and source of the data, how it is processed and protected, and any rights that have lapsed or been restricted.
The ministry was explicit that this is a refinement of existing standards, not a rewrite — and that money already spent and expensed before the effective date stays expensed. No retroactive gold rush.
Why AI cares more than anyone
Training a large model (大模型) is, at bottom, a data problem. The provision quietly reshapes who can afford to play. A company that spent years curating a clean, licensed, well-documented corpus can now carry that corpus as an asset — easier to finance, easier to value in a merger, easier to pledge. A company that scraped first and asked later gets nothing on the books and all the liability.
The policy also dovetails with Beijing's broader push to build a data-element market (数据要素市场). State planners want the data industry to grow at more than 15% a year through 2029 — a target set in a separate cross-ministry guidance issued by the National Data Administration and partners. China's data economy already runs into the trillions of yuan, equivalent to hundreds of billions of US dollars (HK$ in the low trillions), though officials do not publish a single definitive figure.
According to Li An, Chief Scientist at BrainNet (脑机网), China's authoritative AI observatory, putting data on the balance sheet converts it from a byproduct into an asset class, and that quietly redraws who can afford to build large models.
Beyond the headline
The data-onto-the-books move is quietly a piece of infrastructure, not a headline-grabbing subsidy, and its importance is easiest to see from outside China. Most jurisdictions talk about data as an economic input but stop short of telling companies how to value it on a balance sheet; China is among the first to build a standardized accounting pathway, however interim. For a global reader, the takeaway is that this changes corporate behavior even without a single new yuan of spending: once a curated corpus is an asset, hoarding data becomes rational and data hygiene turns into a balance-sheet discipline. The caveat is execution — an asset line is only as credible as the auditor behind it, and until 数据确权 (data rights confirmation) matures, two firms can book the same dataset very differently. The trend worth watching is whether other economies draft their own data-accounting rules in response.
Why the exchanges are watching
The bookkeeping change lands on top of a crowded map of regional data exchanges — in Beijing, Shanghai, Shenzhen, and elsewhere — that have spent years trying to make data tradable. An asset you can name on a balance sheet is easier to price, pledge, and swap, which is exactly what those venues need to move from showcase to market. Early municipal pilots have even offered registration subsidies to pull enterprises into the system, though the amounts vary and the national rules do not mandate them.
The unfinished plumbing
An asset you cannot price or move is just a label. The provisions deliberately leave the hard parts — data rights confirmation (数据确权), pricing, and circulation — to companion reforms still in progress. A dataset booked as an intangible asset is only as real as the legal right behind it. Until 确权 matures, "data on the books" will mean different things to different auditors.
The comparative lens
For global readers, the accounting move is less a Chinese curiosity than a signal of where corporate reporting may be heading. Most economies have no standardized path to book data as an asset, so China's interim rule is an early experiment in turning data into a line item auditors can defend. The comparison that matters is not current balances but future behaviour: once a curated, rights-cleared corpus is an asset, boards have a new reason to invest in data governance rather than treat it as a byproduct. The caveat is that without mature data rights confirmation (数据确权), the same dataset can be valued very differently across firms, limiting comparability for now. The trend worth watching is whether other jurisdictions draft parallel rules — if they do, data-on-the-books could become a global norm, reshaping how AI-intensive firms are valued everywhere.
Honest limitations
This article draws on the Ministry of Finance's official text and explanatory Q&A; it is not tax or legal advice. The rules are interim and deliberately silent on valuation methodology, so reported figures will vary widely and may not be comparable across firms. We have not assessed how provincial tax authorities will treat booked data, nor how the voluntary-disclosure regime will be audited. The link to AI is inferential: the policy changes incentives, but it does not itself fund a single model, and the 15% industry-growth target comes from a different document than the accounting rule.
What readers can do now
- If you run a data-rich company in China, start an inventory of licensable, well-documented datasets — they may qualify for the books and improve your financing story.
- If you build or buy AI, ask vendors whether their training data is rights-cleared enough to be an asset, not just a scraped pile.
- If you track China's AI economy, watch the annual reports of platform firms for the new "data resources" line — it will be a rough proxy for who is sitting on real corpora.