New report forecasts a sharp increase in China's active AI agents

THE ESSENTIALS

IDC and Inspur Information forecast 197 million active agents in China by 2030.
The projection starts from an estimated 4.95 million in 2026.
Agent counts, computing capacity and market spending measure different things.

A steep forecast for task-performing software

A report released by IDC and Inspur Information on September 21 projects that China's active AI agents will rise from 4.95 million in 2026 to 197 million in 2030. The forecast implies roughly a fortyfold increase over four years, or compound annual growth of 151.2%. It is a projection of adoption, not a measured count of systems already operating at that scale.

Reporting on the study also gives a global forecast of 2.216 billion active agents in 2030, up from 79.4 million in 2026. The comparison places China within a much broader expectation that software will move from answering individual prompts to executing tasks through tools and repeated interactions.

The infrastructure assumptions behind the number

The Ministry of Commerce's services-trade information site summarizes the report's separate projections for computing capacity and expenditure. It cites an expected Chinese AI-computing market of $51.5 billion in 2026 and $150.1 billion in 2030. Its capacity forecast reaches 2,576.5 EFLOPS this year and 10,524.3 EFLOPS in 2030.

These figures should remain separate. An agent count measures deployed software units under the study's definition; capacity is a measure of processing capability; market value is spending or revenue within a defined category. Multiplying the agent count by an assumed subscription price would not reconstruct the report's market forecast.

How to read the adoption claim

A single employee or application can use more than one agent, while an agent can perform many tasks. The headline therefore does not imply 197 million distinct paying customers or an equal number of jobs replaced. It also does not reveal how consistently those systems finish work.

The accessible summaries do not provide enough detail to independently reproduce the report's definition of an active agent or its complete forecasting model. The forecast is best read as IDC and Inspur's scenario for a fast-growing workload. The stronger evidence to track over time will be actual task volumes, deployment retention and the spending needed to support them, rather than treating the 2030 endpoint as an observed fact.