NTT Data Signs 900 MW Bloom Energy Fuel Cell Deal for AI Data Centers
Global deployment aims to bypass electric grid bottlenecks with onsite generation
Japanese technology conglomerate NTT Data has entered into a 900-megawatt deployment agreement with Bloom Energy to install solid oxide fuel cells at data center campuses worldwide. The massive procurement highlights how severe grid delays are forcing data center operators to bring their own power generation directly to computing facilities.
Key details
Under the agreement, Bloom Energy will supply NTT Data with its solid oxide fuel cell (SOFC) systems to provide continuous onsite primary and supplemental power. The hardware deliveries are scheduled to begin in 2027, making it Bloom's largest single deployment agreement with any data center developer to date.
NTT Data operates more than 150 data centers across more than 20 countries. Company leadership emphasized that grid interconnect queues can no longer keep pace with client demand for artificial intelligence clusters. Bloom's stationary fuel cell systems—known as Bloom Boxes—produce electricity through an electrochemical reaction without direct combustion, offering modular capacity that can scale alongside facility construction. While the units can technically run on hydrogen or renewable fuels, current deployments predominantly consume natural gas.
Why this matters
The multi-gigawatt pipeline of artificial intelligence infrastructure is colliding head-on with transmission constraints and multi-year utility wait times in major computing hubs. By securing 900 megawatts of behind-the-meter generation, NTT Data can bypass transmission queues that often take five to seven years to deliver high-voltage service.
This deal also demonstrates how AI workload requirements are reshaping power procurement strategies. Because generative model training and large-scale inference require uninterrupted baseline electricity, hyperscalers and colocation providers are increasingly willing to pay a premium for dedicated on-site power rather than wait for traditional grid upgrades.
Context
Bloom Energy has rapidly expanded its footprint across the tech sector as energy shortages mount. The company previously inked agreements with Oracle for up to 2.8 gigawatts of capacity, including deployments at the Project Jupiter complex in New Mexico, as well as deals with Equinix for over 100 megawatts and American Electric Power for up to 1 gigawatt of off-grid capacity.
As regional transmission organizations and municipal utilities struggle to accommodate unprecedented compute loads, the concept of "bring your own power" (BYOP) has shifted from an emergency contingency to standard practice. Hyperscalers and data center operators are contracting directly for natural gas generators, fuel cells, and microgrids to maintain aggressive expansion targets.
Risks and open questions
Despite the speed-to-power advantage of fuel cells, significant logistical and environmental uncertainties remain. The vast majority of solid oxide systems operate on natural gas, which ties data center expansion to gas pipeline infrastructure. In several jurisdictions—such as Oracle's New Mexico site—fuel cell projects have encountered fierce local regulatory resistance over proposed pipeline construction and associated emissions.
Furthermore, relying heavily on natural gas fuel cells complicates corporate net-zero pledges. While fuel cells emit fewer criteria air pollutants than conventional diesel backup generators and avoid combustion soot, they still generate carbon dioxide unless supplied with green hydrogen or paired with carbon capture systems, both of which remain costly and scarce at gigawatt scale.
What happens next
Initial shipments of the 900-megawatt fuel cell order are slated to commence in 2027, with NTT Data expected to announce the first target facilities across its global footprint over the coming year. Regulators and competing cloud operators will be monitoring how smoothly these onsite generation arrays clear local environmental review and pipeline hookups.
Source: Data Center Dynamics Published on AI Usage Global, author: AUG Bot
