Bloom Energy CorpBloom unveiled a new 800V DC-native fuel-cell architecture for AI data centers, cutting capex and energy losses.
Bloom Energy Corporation unveiled a new 800V DC-native fuel-cell power architecture on September 16, designed to supply continuous direct current to the next generation of AI data centers. The solid-oxide fuel cells generate continuous 800V DC power natively rather than producing AC power that must then be converted to DC, potentially removing conversion stages such as transformers and switchgear and lowering both capital cost and energy losses. Bloom claims the technology can cut non-compute capital expenditures for a 1 GW data center by $3.6 billion, or 27%, and lower five-year total cost of ownership by $5.5 billion, or 9%, compared with traditional AC-based infrastructure, though those figures come from the company's own model and depend on site design, energy prices, equipment costs and customer deployment decisions. Bloom pointed to Nvidia's planned adoption of 800V DC architecture beginning with Rubin Ultra and Kyber systems, and said its 2026 Mid-Year Data Center Power Report found data center leaders expect DC-based architectures to account for 58% of new deployments by 2030. The launch comes as Bloom reported its strongest quarter ever in Q2 2026, with revenue up more than 165% year over year and crossing the $1 billion milestone, and raised full-year revenue guidance to $3.9 billion to $4.2 billion, implying 100% growth at the midpoint, along with adjusted EPS guidance of $2.55 to $2.85. The stock has jumped almost 170% since the beginning of 2026, leaving limited room for weaker-than-expected execution, and the company still needs to demonstrate that data-center operators will deploy the technology at scale.
Bloom Energy CorpBloom unveiled a new 800V DC-native fuel-cell architecture for AI data centers, cutting capex and energy losses.
Constellation Energy Corp
NVIDIA CorporationBloom cited Nvidia's planned adoption of 800V DC architecture starting with Rubin Ultra and Kyber systems.