Enteligent has recently published a comprehensive white paper discussing the integration of 800VDC power systems in cutting-edge AI and GPU data centers. The document sheds light on the challenges faced by traditional AC power setups when supporting modern GPU clusters, highlighting the advantages of 800VDC systems coupled with 50VDC server architectures as a more scalable and efficient solution.
Fewer conductors are required, leading to a significant reduction in copper usage by 50–80%, as outlined in the research.
The white paper points out that AI training clusters often surpass 30 kW per rack, underscoring the limitations of conventional AC systems in managing heat and delivering power effectively. High Voltage DC (HVDC) technology allows for higher power densities, paving the way for enhanced efficiency and reliability.
Implementation of 800VDC systems can yield estimated CapEx savings ranging from $4 million to $8 million per 10 MW deployment, along with an 8–12 % decrease in annual energy-related OpEx.
The report also underscores the compatibility with Open Compute Project standards and the industry trend towards DC-native designs. Enteligent positions 800VDC as a solid foundation for AI-centric and large-scale data centers, enabling higher power densities and sustainable operational efficiency.
