In September 2025, one year after the national computing network service platform began trial operation, it released a set of results including a 31-province identifier system and 52 data express routes. This marks China’s computing network moving from dispersed build-out to interconnection.
What data express routes change
The core idea of “data express” is to turn cross-region data transmission into a schedulable infrastructure service. Previously, cross-region data movement often relied on company-built lines or ad hoc transmission over public networks, with high cost and limited controllability. Routed, identified transmission lets data and compute match on demand across a wider area.
For AI companies, the most direct effect is a change in how compute is obtained: training and inference are no longer bound to local resources and can be scheduled across a wider space according to cost and compliance needs. This follows the layout logic of the east-to-west computing project. See our note on building an integrated national computing system.
Compliance questions behind interconnection
The smoother compute and data flow becomes, the more complex compliance becomes. Three questions need to be faced early.
Classification, grading and flow boundaries. Not all data is suitable for cross-region flow. Important data, personal information and core data are constrained by the Data Security Law and the Personal Information Protection Law, so classification and grading must precede cross-region scheduling. See our note on network data security risk assessment.
Compute identifiers and responsibility. Identified management of compute resources means each call may be recorded and traceable. When using public compute, the responsible party at each stage of transmission, storage and computation should be clarified in advance. See our note on data element market allocation.
Contractual arrangements for cross-region scheduling. When procuring cross-region compute services, agreements should specify data custody, deletion, audit and incident liability. See our note on AI commercial contract clauses.
Companion developments to watch
Infrastructure interconnection is usually accompanied by the creation of regulatory interfaces. Compute operation security monitoring and identifier systems mean compute use will gradually become observable. For AI companies, putting compute procurement and data flows into one compliance ledger is more effective than explaining later. See our data compliance service.
Greater public availability of compute will lower entry barriers over time, but it also extends compliance requirements from owned facilities to the public facilities being used.
This article is based on public information released by the national computing network service platform. It is for general reference only and does not constitute legal advice.