Bloom Energy (BE) Stock Rises as 800V DC Fuel Cells Target AI Data Center Power Demand
Key Takeaways
- •Bloom Energy's stock rose 3.77% to $269.13 after the company released a report on 800V DC fuel cell economics for AI data centers.
- •Bloom estimates a 1 GW AI data center could cut non-compute capital costs by $3.6 billion, a 27% reduction under its model.
- •The company projects five-year total ownership costs could drop by $5.5 billion, about 9%, while noting outcomes depend on facility design, equipment expenses, and energy prices.
- •Bloom's solid oxide fuel cells produce DC electricity onsite and deliver it directly to computing equipment, reducing conversion stages, reliance on transformers and switchgear, and overall energy losses.
- •NVIDIA plans to introduce 800V DC architecture with its Rubin Ultra and Kyber rack systems beginning in 2027, and industry participants expect DC designs to account for 58% of new data center deployments by 2030.

TLDR
- Bloom Energy says its 800V DC fuel cells can cut non-compute costs at large AI data centers.
- BE stock rose after the company linked its fuel cells to rising demand for high-density AI power.
- Bloom estimates a 1 GW AI data center could save $3.6 billion in capital costs.
- The company says 800V DC can reduce power conversion equipment and energy losses.
- DC-based architectures could account for 58% of new data center deployments by 2030.
Bloom Energy Corporation (BE) shares rose 3.77% to $269.13 after the company outlined new economics for powering large artificial intelligence data centers. In a newly released report, Bloom said its 800V DC fuel cells could sharply reduce infrastructure spending at high-density computing facilities, as growing demand for dense computing systems continues to reshape power requirements across the data center industry.
Bloom Energy Targets Lower AI Data Center Power Costs
Bloom Energy's report examines how direct-current power could reduce the infrastructure requirements of large computing facilities. According to the company's economic model, a 1 GW AI data center could cut non-compute capital costs by $3.6 billion — a 27% reduction under the assumptions included in the analysis. Non-compute costs cover the infrastructure surrounding the computing hardware itself: power distribution, conversion equipment, and the electrical systems that feed each rack.
The company also highlighted potential savings over longer operating periods, estimating that its approach could reduce five-year total ownership costs by $5.5 billion, or roughly 9%. Bloom cautioned, however, that project outcomes depend on facility design, equipment expenses, energy prices, and other site-specific conditions.
The company generates DC electricity onsite through solid oxide fuel cells and delivers that power directly to computing equipment. This structure removes several conversion stages that conventional alternating-current systems require before electricity reaches computing racks. As a result, Bloom says operators can reduce equipment needs while limiting energy losses across the electrical system.
800V DC Architecture Addresses Rising Power Density
Modern data centers traditionally receive alternating-current electricity from grids, turbines, or onsite engines before converting it into direct current, since computing hardware itself runs on DC. High-density computing racks, however, now require larger power loads within smaller physical spaces.
Next-generation racks increasingly call for 800V DC inputs because lower-voltage systems face limits under heavier power demand. Moving the same amount of power at higher voltage means lower current, which limits resistive losses and reduces the conductor volume needed to route electricity through a facility. NVIDIA plans to introduce 800V DC architecture with its Rubin Ultra and Kyber rack systems beginning in 2027. As computing companies increase rack density, power providers must adapt their infrastructure accordingly.
Bloom's fuel cells generate continuous DC electricity directly, avoiding the initial AC-to-DC conversion used in conventional systems. The design can also reduce reliance on transformers, switchgear, and other electrical components that carry extended delivery schedules — a point Bloom tied to the emerging requirements of next-generation data centers.
Data Center Expansion Supports Bloom Energy Strategy
Data center operators continue to seek additional power as computing workloads increase electricity consumption across major markets. Grid connection delays and equipment shortages have also encouraged developers to examine onsite generation systems.
Bloom previously reported that industry participants expect DC architectures to capture 58% of new data center deployments by 2030. The company sees onsite DC generation as one method for reducing electrical complexity within large computing facilities. Its strategy also targets lower copper requirements, because direct-current systems can remove some traditional electrical equipment.
The broader shift gives Bloom an opportunity to position its fuel cells alongside changing data center electrical standards. Still, actual savings will vary, as each project faces different construction, equipment, and energy costs. With NVIDIA's 800V DC introduction slated for 2027 and DC designs expected to capture 58% of new deployments by 2030, Bloom's projections will have clear reference points as next-generation facilities come online. BE stock ended higher after the company presented 800V DC onsite generation as a solution for the expanding data center power market.
Source: Blockonomi