China's AI Token Consumption to Hit 100 Quadrillion by 2026, Report Forecasts
A report from the China Telecom Research Institute projects China's annual AI token consumption will reach 100 quadrillion in 2026 and exceed 35,000 quadrillion by 2030. This explosive growth represents an annual compound growth rate of nearly 12x as the industry shifts toward AI agents and commercial application deployment. The forecast highlights a fundamental transition in the AI infrastructure landscape from pre-training models to massive inference computing. As autonomous AI agents proliferate, computing demands will shift heavily toward serving real-time usage, with inference expected to capture 80% of China's AI compute market by 2029. China's annual compute demand is expected to increase by nearly 10x per year over the next two to three years, driven by leading tech companies spending nearly 600 billion yuan (~$84 billion) on AI capital expenditures in 2026. On a global scale, the report expects active AI agents to expand from nearly 80 million in 2026 to 2.216 billion by 2030.
## BACKGROUND
Tokens are the basic units of text or character fragments processed by Large Language Models (LLMs) during both input and response generation. AI agents leverage LLMs to perform complex, multi-step autonomous workflows, which continuously process context and invoke tool calls, consuming vastly more tokens during inference compared to basic single-turn chat queries.