- The project aims to integrate renewable power, AI computing and token production into a single infrastructure platform
- It reflects China’s push to make energy efficiency—not just computing power—a core metric in the AI race
China’s first city-scale green token factory will be built in Changzhou under a partnership signed during the 2026 World Artificial Intelligence Conference (WAIC), marking a new attempt to integrate renewable energy generation with AI computing infrastructure.
The project, anchored by a joint agreement reached on July 18, brings together seven partners, including the Changzhou Data Bureau, Changzhou National Hi-Tech District, State Power Investment Corp., Alibaba Cloud, Wangsu Science & Technology and DeepEngine.
Once completed, the facility is expected to produce up to 60 trillion AI tokens annually.
From model race to token race
The project reflects a broader shift in the AI industry from competing on model capabilities to competing on the efficient production of AI inference.
Tokens—the basic units processed by large language models—have emerged as a key measure of AI consumption. Every chatbot query, code generation request or AI agent task consumes tokens.
As demand for inference surges, the industry’s focus is increasingly shifting from how many chips a data center owns to how many high-quality tokens it can generate per unit of electricity.
The Changzhou project aims to raise computing utilization from the industry norm of 30%-40% to more than 65%, increasing effective token output by an estimated 62.5%.
A new model for AI infrastructure
Rather than simply colocating data centers next to power plants, the project seeks to tightly integrate renewable electricity generation, grid scheduling and AI computing into a unified system.
The facility plans to deploy technologies including direct renewable power connections, 800-volt high-voltage direct current (HVDC) transmission, solid-state transformers, liquid cooling and AI-based energy management.
The goal is to address one of the AI industry’s biggest challenges: sustaining exponential growth in computing demand without allowing energy costs to rise at the same pace.
Why it matters for global readers
For international observers, the Changzhou project offers an early example of China’s emerging “power-compute integration” strategy, in which energy infrastructure and AI infrastructure are designed as a single system rather than as separate industries.
Changzhou, one of China’s leading new energy manufacturing hubs, is already home to several national pilot programs in low-carbon energy and smart power systems, making it a natural testing ground for the model.
Chinese policymakers envision expanding the approach across the Yangtze River Delta and eventually linking low-cost renewable energy resources in western China with AI computing demand at home and abroad.
If successful, token factories could become not only AI infrastructure, but also a new export model combining energy, computing capacity and digital services.
