How solid-oxide fuel cells power a data center
How onsite solid-oxide fuel cells make electricity, why natural gas still brings emissions, and which grid constraints remain.
A fuel cell beside a data center can make electricity without waiting for every electron to arrive through a distant power plant and transmission line. That makes the technology attractive when a campus wants power faster than the local grid can expand.
But “fuel cell” can sound cleaner and simpler than the complete system really is. A stationary fuel cell still needs a steady fuel supply, electrical equipment, permits, maintenance and a plan for failures. If it runs on natural gas, it still releases carbon dioxide.
The useful question is not whether fuel cells are good or bad. It is what they do differently from combustion generators and the grid—and which constraints remain after the equipment reaches the site.
Electricity from chemistry instead of a flame
A fuel cell converts chemical energy directly into electricity. The Department of Energy explains that a fuel cell has two electrodes around an electrolyte. Charged particles move through the electrolyte while electrons travel through an external circuit, creating usable electric current. Fuel and oxygen must keep flowing for the device to keep producing power. DOE: Fuel cells
That is different from burning fuel to make hot gas, using the gas to spin a turbine and using the turbine to turn a generator. The fuel cell skips the flame and moving generator step. It is more like a continually supplied battery than a stored battery: the electrochemical reaction produces electricity, but the fuel is delivered from outside.
The analogy has a limit. A data-center installation is not one sealed box. Individual cells are grouped into stacks; stacks sit inside modules; modules are combined with fuel-processing, air, power-conversion, cooling and control equipment. The output still has to be conditioned and connected to the campus electrical system.
Why solid oxide matters
The systems in Bloom Energy’s November agreement with NTT DATA use solid-oxide fuel cells. “Solid oxide” describes the ceramic electrolyte inside the cell. These cells operate at high temperature, which lets them use several fuels and convert them efficiently, but the heat also shapes how they start, stop and follow changing demand.
A DOE technology assessment placed solid-oxide systems in a broad operating range of 600–1,000 degrees Celsius. It reported electrical efficiencies of roughly 50% to the low 60% range for the technology available at the time. Those are technology-assessment ranges, not a performance guarantee for a new installation. DOE: Stationary Fuel Cells Technology Assessment
High temperature also creates a practical constraint. DOE described solid-oxide systems as less suited to rapid load following because repeated heating and cooling can damage stacks. A data center with sharply changing demand may therefore need controls, batteries, the grid or other equipment to handle fast changes. The fuel cells can provide steady output while another resource absorbs the quickest swings.
The fuel determines the carbon story
Hydrogen entering a fuel cell can produce electricity and water at the point of use. Many stationary systems instead receive natural gas or biogas. They convert that fuel into a hydrogen-rich stream inside or near the system before the electrochemical reaction.
The Environmental Protection Agency’s fuel-cell technology guide explains the distinction clearly: avoiding combustion greatly reduces local emissions of common air pollutants, but a natural-gas-fueled system still produces carbon dioxide during fuel processing and the cell reaction. It also depends on the upstream gas supply. EPA: Fuel-cell technology characterization
That means “noncombustion” and “zero carbon” are not synonyms. Natural-gas fuel cells can have lower local nitrogen-oxide and particulate emissions than combustion generators, yet their climate effect depends on operating efficiency, fuel production and transport, methane leakage, and what electricity they replace. Hydrogen changes the calculation only after its production and delivery are included.
Why data centers want power at the fence
Onsite generation can shorten the physical path between electricity production and servers. It may let a developer add capacity in modules while a larger utility connection is still being studied or built. The arrangement can also support a microgrid that separates from the wider grid during some disturbances.
Those benefits explain the scale of the latest proposal. On November 5, NTT DATA and Bloom Energy announced an agreement covering 900 megawatts of planned onsite generation across multiple data-center projects, with initial shipments scheduled for 2027. The companies did not name the sites, divide the capacity among them or disclose the fuel and grid arrangements. The 900 MW figure describes the agreement’s planned total, not operating generation today. Bloom Energy: NTT DATA agreement
Modularity can speed equipment delivery, but it does not eliminate site work. A project still needs foundations, gas or hydrogen infrastructure, switchgear, transformers, protection systems and commissioning. Local authorities may need to review land use, air emissions, fire safety and fuel connections. The exact list depends on the location and design.
Reliable equipment is not the same as a reliable service
A fuel cell can generate continuously as long as fuel, air and the supporting equipment remain available. That “as long as” contains the reliability problem. Gas delivery can be interrupted. A module can need maintenance. Power electronics or campus distribution can fail even while the cell stacks are healthy.
Operators address those risks with multiple modules, reserve capacity, grid service, storage or other generators. Each layer covers particular failures for a particular time. A campus claiming resilient onsite power should therefore explain which loads the system supports, what happens when a module is offline, how much fuel supply is firm and whether the grid remains part of the backup plan.
There is also a difference between primary and standby power. Equipment designed to run around the clock faces different fuel, maintenance and emissions questions from emergency generators tested occasionally. Public documents should say which role the system will play.
Water and heat need site-specific answers
Fuel cells produce heat as well as electricity. In some buildings, useful heat can be recovered for hot water or another thermal load, improving total system efficiency. A data center may not have a large, constant need for that kind of heat, so combined-heat-and-power figures should not be assumed to apply automatically.
Bloom says its technology uses minimal water during normal operation and produces low local air emissions. Those are manufacturer claims about its platform. A project-level review still needs the chosen configuration, fuel treatment, cooling design and local permit information. The wider data center has its own cooling and water requirements, separate from the generating equipment.
Noise also belongs in the site review. Fuel cells have no combustion turbine inside the stack, but a complete plant includes blowers, pumps, compressors and electrical gear. The relevant evidence is a project’s acoustic design and permit, not the absence of a flame.
How to read the next fuel-cell announcement
Start with status. Is the document a memorandum, a framework agreement, a firm equipment order, a permitted project, a commissioned system or measured operation? Then find the capacity boundary: one site, an initial phase or a total spread across future projects.
Next ask what fuel is planned and how it reaches the site. Natural gas, biogas and hydrogen carry different infrastructure and emissions questions. Ask whether the installation supplies normal load, backup load or both, and what supports the campus during maintenance or fast demand changes.
Finally, separate the stack from the full power system. Fuel cells can move generation closer to the servers and reduce dependence on a delayed grid expansion. They cannot replace every wire, transformer, control, fuel contract or permit around them.
The technology’s value is specific: steady, modular electricity made onsite through an electrochemical process. The proof for any announced campus will be equally specific—named sites, approved designs, delivered equipment, operating fuel arrangements and measured performance.