With the development of artificial intelligence (AI) technologies, the number of global data centers and high-performance computing (HPC) facilities is rapidly increasing. AI computing, particularly large language model training and real-time inference services, places unprecedented demands on electricity supply. Traditional power grids face load fluctuations and peak pressures, and the continuous expansion of AI data centers is driving global power infrastructure into a new phase.

Challenges of AI Computing for Power Grids

Compared with traditional servers, AI data centers have distinct characteristics:

• High power demand: Each GPU rack may consume tens of kilowatts, and multiple racks operating simultaneously can reach total power usage in the megawatt range.

• Continuous workload: AI services require 24/7 operation, with power demand spiking during peak periods.

• Cooling and heat management: High-density GPU clusters generate substantial heat, necessitating advanced liquid-cooled or air-cooled systems to maintain stability.

These factors make AI data centers energy-intensive users, and load fluctuations can directly affect grid stability. Voltage fluctuations or outages in some regions may threaten the reliable operation of AI computing workloads.

Role and Value of Energy Storage Systems

To address these challenges, Battery Energy Storage Systems (BESS) are increasingly adopted by AI data centers. Key functions include:

• Grid stability support: Energy storage can immediately release power during grid fluctuations, ensuring uninterrupted AI computing operations.

• Peak shaving: Energy storage provides additional power during peak loads and stores energy during low-demand periods, optimizing energy efficiency.

• Backup power: In the event of sudden outages or grid failures, energy storage ensures stable power for critical workloads.

• Renewable energy integration: When using solar or wind power, energy storage stores excess electricity and releases it during high-demand periods, balancing supply and demand.

Dagong ESS Products in AI Power Infrastructure

Different scales of data centers can deploy different Dagong ESS solutions:

1. Medium commercial energy storage

• 100kWh–144kWh Air-Cooled ESS

• 241kWh–418kWh Air-Cooled ESS
  Modular design allows flexible expansion, suitable for small to medium AI data centers.

2. High-density liquid-cooled storage

• 241kWh Liquid-Cooled ESS

• 372kWh Liquid-Cooled ESS
  Liquid cooling maintains stable temperature in high-power environments, enhancing efficiency and lifespan.

3. Large-scale containerized storage

• 5MWh Liquid-Cooled ESS Container
  Suitable for campus-level or regional AI computing infrastructure, supporting real-time monitoring, fault logging, and capacity expansion.

Dagong ESS systems feature 8000+ cycle life and 15+ years of reliability, ensuring stable operation in high-load AI data centers.

Global Trends and Future Outlook

AI computing is reshaping the global energy landscape:

• Increasing electricity demand: AI applications across finance, healthcare, research, and manufacturing will significantly raise global power consumption.

• Rising storage demand: Energy storage addresses peak loads, improves energy efficiency, and reduces carbon emissions.

• Grid modernization: The expansion of AI data centers drives grid upgrades, introducing smart controls and distributed storage systems.

Over the next 5–10 years, AI computing and energy storage systems will become critical forces driving upgrades in power infrastructure, providing reliable energy support for global intelligent computing.