Bloom Energy Has Introduced AI Power Architecture
The new 800V DC-native system aims to streamline electricity delivery for intensive data center operations.
Updated on Sept. 20, 2026 in Data Centers
Live Poll
Will efficient DC power architectures lower the long-term operational costs of AI data centers?
Bloom Energy has released an 800V DC-native fuel-cell architecture designed to provide on-site power for AI data centers. By generating electricity through solid oxide fuel cells that run on natural gas, biogas, or hydrogen, the system allows facilities to bypass traditional AC grid infrastructure.
Why it matters
This architecture matches the direct power input requirements of modern AI accelerators, enabling providers to eliminate inefficient electrical conversion steps. The shift is expected to significantly lower costs and power complexity for massive computing clusters.
The 800V DC architecture supports a projected 27% reduction in non-compute capital expenditures for a 1 GW data center. DC-based designs are projected to capture 58% of all new data center deployments by 2030.
The players
Bloom Energy
This company specializes in solid oxide fuel cell technology used for on-site power generation.
NVIDIA
A global leader in graphics processing units and hardware essential for AI development and data centers.
Nebius
An international technology firm that recently contracted Bloom Energy for a 328 MW power supply project.
Oracle
A multinational computer technology corporation that maintains a partnership with Bloom Energy.
Equinix
A digital infrastructure company that operates data centers and utilizes power solutions from Bloom Energy.
The details
The system delivers continuous DC output directly to AI hardware, optimizing efficiency by removing the need for standard AC conversion. Nebius has already contracted Bloom Energy for a 328 MW power supply project, while industry leaders including Oracle, Equinix, and Brookfield maintain active partnerships with the company.
Timeline
NVIDIA plans to adopt 800V DC rack standards starting in 2027.
DC-based architectures are projected to reach 58% market share by 2030.
The Tech Race
This move signals a major shift in the data center industry as providers pivot from legacy AC infrastructure toward more efficient, native DC-based power distribution systems. This trend directly positions the technology to support the massive energy requirements of AI hardware manufacturers like NVIDIA.
For users and developers, this technology promises to enable more stable and efficient AI services by reducing the cooling and power bottlenecks common in current data centers. While the direct financial impact applies to infrastructure owners, it creates a more sustainable foundation for the software and AI platforms consumers use daily.
The takeaway
The industry-wide move toward 800V DC power architectures highlights the extreme energy demands of modern artificial intelligence. Adopting these streamlined systems could be the deciding factor in making large-scale AI deployment economically viable over the coming decade.
Further reading
Learn more about the latest infrastructure trends in the Data Centers section.
Live Poll
Will efficient DC power architectures lower the long-term operational costs of AI data centers?
