Enteligent has released a white paper detailing the adoption of 800VDC power systems for next-generation AI and GPU data centres. The report highlights how traditional AC power paths face economic and reliability challenges when supporting modern GPU clusters, while 800VDC systems paired with 50VDC server architectures provide a more scalable and efficient solution.
Fewer conductors are needed, with copper requirements reduced by 50–80%, according to the findings.
The paper notes that AI training clusters often exceed 30 kW per rack, revealing the limitations of legacy AC systems in thermal management and power delivery. High Voltage DC (HVDC) enables higher power densities, supporting efficiency and reliability improvements.
Adopting 800VDC systems can result in projected CapEx savings of $4 million to $8 million per 10 MW build and an 8–12 % reduction in annual energy-related OpEx.
The report also emphasises alignment with Open Compute Project standards and the industry shift toward DC-native architectures. Enteligent positions 800VDC as a suitable foundation for AI-first and hyperscale data centres, supporting higher power densities and long-term operational efficiency.
