SenseCore Jointly Launches "Computing-Power-Electricity Synergy Platform" to Build Green AI Data Center "Model Demonstration Facility"

During the 2025 World Artificial Intelligence Conference (WAIC), SenseCore, in collaboration with DaMao AI and CATL Capital (CATL's exclusive industrial investment platform), launched the "Lingang AIDC Computing-Power-Electricity Synergy Platform". The platform was first implemented at SenseTime’s Lingang AIDC, China's first 5A-rated intelligent computing facility. By establishing intelligent infrastructure that deeply integrates artificial intelligence with energy systems, the initiative propels the Lingang AIDC into a new phase characterized by "low-cost, high-efficiency, and sustainability", while providing replicable and scalable solutions for greening AI infrastructure upgrades.

Surge in AI Computing Demand Meets Energy Bottlenecks, Computing-Power-Electricity Synergy Emerges as Key Solution

The AI era has witnessed an exponential surge in computing power demand, yet challenges in energy structure transformation, supply-demand balance, and carbon emission constraints have become critical bottlenecks for the development of intelligent data centers. The key to achieving the green transformation of intelligent infrastructure lies in establishing a synergistic development model where "computing power and electricity sync and coexist harmoniously." This is accomplished through a dual-path approach, by "leveraging electricity to drive greener computing power upgrades" and "utilizing computing power to enable smarter electricity system transformation".

The SenseTime Lingang AIDC Computing-Power-Electricity Synergy Platform embodies this practice of mutual empowerment between computing and energy systems. It integrates SenseTime's self-developed multimodal large models, DaMao AI's energy-specific large models, and CATL's energy storage technologies. By combining heterogeneous computing power elastic scheduling platforms with dynamic flexible energy dispatch platforms, and leveraging comprehensive data from SenseTime's intelligent operation system, it creates an end-to-end intelligent scheduling system covering "data-algorithms-strategies-execution-verification." Through its dual application loop of "computing adapts to electricity availability" and "electricity adapts to computing needs," the platform achieves precise matching between computing resources and energy supply.

Self-Developed Energy Large Model + Smart O&M Saves 3 Million kWh Annually SenseCore's energy-optimized large model system, trained on massive industry knowledge texts, structured energy data, and computing power metrics - enhanced by domain-specific knowledge bases - has established an LM+Agent architecture that significantly improves professional accuracy and applicability across sectors. Through this model's integration with smart facility operations, the platform achieved an annual PUE optimization below 1.28, saving 3 million kWh of electricity annually.

Innovative "Energy Block" Algorithm Boosts Predictive Accuracy: The system binds fundamental energy characteristics, user consumption patterns, and power balance rules with computing servers to create basic "Energy Block" TOKENs. By continuously refining the algorithm architecture through multiple rounds of pretraining, it delivers superior prediction precision and generalization capability for complex scenarios.

15-Minute Dynamic Scheduling Achieves Optimal Energy Balance: The platform performs high-frequency predictions at 15-minute intervals, with decision-making optimizations iterated every 5 minutes. By analyzing real-time energy status and load forecasts, it automatically generates optimal scheduling strategies. Through cross-system coordinated execution, the system achieves precise computing power prediction, load forecasting, and dynamic strategy generation/refinement.

Currently, the Energy Block-based model has achieved over 88% accuracy in energy demand prediction and surpassed 93% accuracy in decision-making, with projections indicating it will reach industry-leading levels of 90-95% prediction accuracy and 95%+ decision accuracy by year-end through the continuous optimization of algorithms and energy storage systems.

Collaborating with the Industrial Chain Ecosystem, Replicating the Model Across China Through Benchmark Practices

SenseCore is committed to building a future-oriented green AI foundation through cutting-edge technological approaches. As China's largest single-site intelligent computing center, the SenseTime Lingang AIDC was certified as the nation's first 5A-rated intelligent computing facility in September 2024, with its operational expertise now supporting SenseTime's nationwide computing power network deployment. The tripartite collaboration in launching the Lingang AIDC Computing-Power-Electricity Synergy Platform represents a significant technological advancement for green data centers, while establishing a replicable model of "compute-energy synergy" for intelligent computing center development across China.

"Energy large models serve as the core engine driving intelligent transformation in energy systems," said Jian Yumin, CEO of DaMao AI. "The platform model we co-developed with SenseTime has achieved high generalization capability and dynamic responsiveness. Moving forward, we will accelerate its deployment across more AI infrastructure nodes nationwide to realize integrated smart control from energy storage-load scheduling to source-grid-load-storage coordination. Through unified scheduling and trading services, we aim to empower the AI industry's development and facilitate the achievement of dual-carbon strategic goals."

Yi Junyuan, Executive Director of CATL's industrial investment fund CATL Capital, said, "Intelligent computing centers represent crucial loads in new-type power systems, with energy storage serving as the most critical regulation resource. Meanwhile, AI models and algorithms form the foundation for system regulation and scheduling. Therefore, computing-power-electricity synergy and integration represent a definitive industry trend and a key focus of our strategic investments. This collaboration will jointly create a benchmark case in the 'AI+Energy' field, with future explorations into technological innovations combining renewable energy and computing infrastructure."

Lin Hai, General Manager of Intelligent Computing Centers at SenseCore Business Group, said, "The green transformation of AIDCs has become a pressing issue to tackle. What SenseCore has developed is far more than a singular technological platform - it constitutes a comprehensive system integrating core competencies and practical methodologies that deeply converge AI with energy technologies. We are establishing an industry benchmark for green AI infrastructure that combines universal applicability with sustainability, enabling every computing node to acquire cost-effective green capabilities and ultimately drive the sustainable development of AI infrastructure."