As the power density of AI computing clusters continues to increase, traditional power supply architectures face challenges such as high power, high density, and rapid power fluctuations. NVIDIA is driving the evolution of next-generation AI data centers (AI Factories) toward an 800V HVDC (High-Voltage Direct Current) power supply architecture and is promoting the establishment of a corresponding integration and certification system for energy storage systems. Its core objective is to address the power supply bottlenecks—including high power, rapid fluctuations, and high density—caused by AI computing clusters through a “high-voltage direct current + energy storage” solution.

By increasing the supply voltage, 800V HVDC reduces the transmission current for the same power output, thereby minimizing line losses and alleviating pressure on the power distribution system, and supporting the construction of AI computing infrastructure with higher power density. At the same time, energy storage systems enhance the AI data center power supply system’s resilience to load fluctuations and grid anomalies through capabilities such as rapid power response, grid-forming operation, and coordinated on-grid and off-grid operation.

In an 800V HVDC architecture, energy storage systems not only serve as backup power sources in the traditional sense but must also provide stronger dynamic power support and grid adaptability, including:

For high-power-density AI data centers, future energy storage systems will work in close coordination with 800V HVDC architectures to form a layered energy architecture combining “short-term dynamic support” and “long-term energy storage”: front-end rapid-response devices such as supercapacitors can suppress millisecond-level power fluctuations, while battery energy storage systems handle power balancing and backup power supply over longer time scales.