NTT DATA Commits 900 MW to Bloom Energy Fuel Cells for AI Data Center Primary Power
Bloom's fuel cells offer 90-day deployment against a five-year grid interconnection wait
NTT Global Data Centers has agreed to deploy 900 megawatts of Bloom Energy solid oxide fuel cells across its data center portfolio worldwide, treating the technology not as backup generation but as the primary power source for AI workloads. The agreement, announced November 5, 2026, represents Bloom Energy's largest international project to date and marks a meaningful shift in how major Tier-1 data center operators are thinking about power sourcing: when the utility grid cannot deliver capacity fast enough to match AI demand, bypass it entirely.
"Our clients are building AI capabilities at a pace the grid alone cannot always match, and speed to power has become one of the defining factors in delivering capacity when and where it's needed," NTT Global Data Centers CEO Doug Adams said in the announcement.
The Grid Interconnection Constraint Driving Behind-the-Meter Power
The strategic logic behind the deal is inseparable from one of the most significant infrastructure bottlenecks in the AI buildout: transmission interconnection queues. Lawrence Berkeley National Laboratory's Queued Up 2026 Edition found that the median wait time for a new power plant to complete the grid interconnection process had grown from 22 months in 2008 to 61 months for plants finished in 2025. That is more than five years — an eternity for data center operators whose hyperscaler clients are committing to AI capacity years in advance.
The International Energy Agency projects global data center electricity consumption will roughly double from approximately 485 terawatt-hours in 2025 to 950 terawatt-hours by 2030. NTT DATA's own "Can Data Centers Keep Pace with AI? A Global Data Center Outlook" report projected 23 to 30 percent annual demand growth through the end of the decade, while flagging power availability, equipment supply, land, and labor as the factors most likely to constrain that growth.
Against that backdrop, a technology that can be deployed in roughly 90 days — Bloom says it delivered a completed installation to Oracle in just 55 days, ahead of a 90-day schedule — offers a qualitatively different competitive position than a grid connection that may not come online until 2031.
How Solid Oxide Fuel Cells Generate Primary Power Without Combustion
Bloom Energy's fuel cell technology differs from both diesel generators and conventional power plants in a way that matters for dense urban or industrial sites. Rather than burning fuel, solid oxide fuel cells convert it electrochemically. The system heats a ceramic electrolyte to between 700 and 1,000 degrees Celsius, at which temperature oxygen ions migrate from the cathode to the anode, where they react with hydrogen or reformate gas derived from natural gas or biogas. The result is a direct electrical current, water vapor, and concentrated carbon dioxide — but no combustion, and therefore near-zero nitrogen oxides and sulfur oxides.
That chemistry produces meaningful efficiency advantages. Running on hydrogen, Bloom's systems reach approximately 60 percent electrical efficiency, compared to roughly 35 to 50 percent for natural gas combined-cycle plants depending on scale. When waste heat is recovered for data center thermal management, combined efficiency can approach 90 percent. The modular architecture — individual fuel cell units stack into servers, servers aggregate into larger arrays — allows incremental deployment timed to data center construction phases rather than a single all-or-nothing installation event.
The technology's power density is also notable for land-constrained markets: Bloom claims approximately 100 megawatts per acre.
There are, however, genuine constraints that the announcement does not address. Bloom's current deployments primarily run on natural gas, which means they still produce carbon dioxide at the point of generation, even without combustion. The concentrated CO₂ output is theoretically suited to carbon capture, but no commercial carbon capture deployment at Bloom's data center scale is currently operational. For operators with public net-zero commitments, natural-gas fuel cells represent a lower-emission bridge rather than a zero-carbon endpoint. Full decarbonization would require either hydrogen supply chains that do not yet exist at scale or biogas sourcing that cannot supply gigawatt-class deployments.
Community opposition, while less severe than for large gas turbine installations, remains a practical consideration for siting new generation capacity at scale. Adams indicated that community acceptance was as important to NTT DATA as the megawatts of capacity the installations deliver.
Oracle Already on the Same Path; Bloom's Revenue Surges
NTT DATA is not pioneering an untested strategy. Oracle committed to up to 2.8 gigawatts of Bloom fuel cell capacity — with 1.2 gigawatts already under contract — as part of its Project Jupiter AI infrastructure expansion, switching from natural gas turbines to fuel cells partly to reduce community opposition at proposed sites. Bloom's fuel cells produce approximately 92 percent fewer nitrogen oxides than conventional gas turbines, a meaningful advantage in permitting at urban and suburban locations. AEP agreed to source up to 1 gigawatt of Bloom-generated power for data center customers in the same period. Equinix and Nebius have also announced Bloom deployments at meaningful scale.
Collectively, Oracle and NTT DATA alone now account for announced Bloom commitments of approximately 3.7 gigawatts — more than double Bloom's estimated cumulative global installed base of approximately 1.4 gigawatts as of early 2026. The pipeline has fundamentally outpaced the installed fleet. Bloom says it is on track to reach 2 gigawatts per year of manufacturing capacity by the end of 2026, up from roughly 1 gigawatt today, with US-based production facilities providing supply chain certainty for domestic deployments.
The financial results reflect this trajectory. Bloom reported Q3 2026 revenue of $1.24 billion, a 138.1 percent year-over-year increase, and raised full-year 2026 guidance to between $4.4 and $4.6 billion. Approximately 97 percent of Q3 revenue came from domestic US deployments — the NTT DATA deal signals an explicit push to convert the company's internationally recognized technology into international contracts at comparable scale.
Bloom's principal competitor in stationary fuel cells, FuelCell Energy, reported Q3 fiscal 2026 revenue of approximately $33 million on a declining trajectory, suggesting the consolidation of data center demand around Bloom's platform rather than distributing it across multiple suppliers.
A Stock Drop Contradicts the Deal's Bullish Signals
The market's reaction on November 5 did not match the announcement's strategic weight: Bloom Energy shares fell approximately 14 percent to around $171.52 from a prior close near $200.40.
The divergence reflects political risk layered onto the operational story. Democratic gains in the November 3 midterm elections raised investor concern about federal AI policy that could restrict or slow data center approvals, and in Texas, Governor Greg Abbott had issued a moratorium on new large-scale data center connections to the state grid in August 2026. Incoming Governor Gina Hinojosa, who won the November 3 governor's race, had signaled support for even stricter data center restrictions — a significant concern given Texas's prominence as a data center market and the near-total domestic concentration of Bloom's current revenue base.
Whether those political headwinds materially slow the execution of already-signed agreements remains to be seen. NTT DATA's 900 MW commitment is spread across global data center projects, reducing single-state regulatory exposure. The initial shipments are expected to begin in 2027. No financial terms were disclosed.
Behind-the-Meter Primary Power as a Structural AI Strategy
The combined Oracle and NTT DATA commitments are large enough to suggest that behind-the-meter primary power has crossed from an opportunistic workaround into a structural layer of the AI data center stack. When two of the world's largest data center operators independently reach the same conclusion — that onsite fuel cell generation is preferable to waiting in a five-year utility queue — it creates a model that other operators must either replicate or explain why they have not.
The strategic question for the industry over the next 24 to 36 months is whether the carbon-emission profile of natural-gas-fired fuel cells can be addressed before regulatory or reputational pressure forces a correction. Bloom's roadmap toward hydrogen compatibility exists, but hydrogen supply infrastructure at gigawatt scale does not. If data center operators continue committing gigawatts of natural gas fuel cell capacity while making net-zero pledges for 2030 or 2035, the gap between those two commitments will require a credible explanation. For NTT DATA, Oracle, and Bloom Energy, the more immediate test is whether the 90-day deployment advantage holds as projects scale from hundreds to thousands of megawatts — and whether the interconnection queues that created the opening for fuel cells begin to shorten, or continue to grow.