Sensetime's Lin Hai, General Manager of the Intelligent Computing Center under the company's Large-Scale AI Infrastructure Division, recently highlighted that global computing infrastructure is entering a new phase of strategic competition focused on "computing efficiency." As a leading domestic AI infrastructure provider, Sensetime's Large-Scale AI Infrastructure has developed a systematic approach to integrating computing and energy resources, presenting a Chinese model for "computing-energy" synergy. This initiative aims to secure a technological advantage in the next stage of AI infrastructure competition.

In July 2025, Sensetime launched its independently developed "Intelligent Computing-Energy Coordination Platform," leveraging long-term operational experience from its 10,000-GPU AI Data Center (AIDC). The platform integrates computing and electricity data, delivering notable economic and social benefits in real-world applications.

Sensetime's proprietary energy AI model employs a multimodal Mixture of Experts (MoE) architecture, trained on vast datasets including industry knowledge, structured energy data, and computing performance metrics. It predicts computing load trends and dynamically optimizes resource allocation by factoring in electricity pricing, renewable energy ratios, energy storage status, and grid demands. This transforms data centers from "fixed loads" into "adjustable resources," enabling proactive "computing follows electricity, electricity follows computing" scheduling.

Currently, the model achieves over 88% accuracy in energy demand forecasting and over 93% in decision-making. With ongoing algorithm and energy storage optimizations, forecasting accuracy is expected to reach industry-leading levels of 90%-95%, while decision-making accuracy will exceed 95%.