Goldman Sachs estimates fuel cells could supply 6–15% of incremental data center power demand by 2030 — roughly 8–20 GW of new capacity, because new gas turbine projects now face 5-plus-year lead times while modular fuel cell systems can be deployed in under 12 months. With grid interconnection queues stretching 3–7 years in many U.S. markets, the procurement question has shifted from emissions to schedule: which megawatts can be online before the AI workload ships.
One caveat governs everything below. The performance data in this category — efficiency, emissions, noise, availability, deployment speed — originates almost entirely with fuel cell manufacturers or with content they sponsor. Independent, buy-side measurement is thin. Every vendor figure here is labeled as such. The contract capacities are the sturdiest class of claim in the set, because they were announced by public companies on both sides of the transaction — but they are still press releases, not audited disclosures.
Key Takeaways
- Fuel cells deploy in under 12 months against 5-plus years for new gas turbine projects. Bloom says it delivered Oracle a fully operational system in 55 days.
- Installed capex runs $3,000–$5,000/kW against $800–$1,500/kW for simple-cycle gas turbines and $300–$700/kW for diesel gensets. The case is built on schedule and fuel efficiency, not sticker price.
- The public deployment record is real but concentrated in one vendor. Bloom holds the AEP, Equinix, and Oracle contracts; FuelCell Energy is scaling a packaged 12.5 MW block.
- Natural gas is the commercial default, not hydrogen. Fuel cells move your dependency from the electric utility to the gas utility rather than eliminating it.
- Stack replacement every 5–10 years is a mid-life capital event with no equivalent on a gas turbine at the same interval. It belongs in the term sheet, not the footnotes.
Why Fuel Cells Are Suddenly on the Data Center Procurement Short-List#
For a hyperscaler or colocation developer staring at a stranded shell — racks ordered, GPUs landed, utility service date slipping — the argument is no longer environmental. It is that a fuel cell block can be energized inside the window where a gas turbine order is still in the OEM queue.
The 2024–2026 contract record makes that explicit. Bloom has signed agreements with AEP for up to 1 GW with an initial 100 MW order, Equinix exceeding 100 MW across 19 IBX data centers, and Oracle for up to 2.8 GW with an initial 1.2 GW contracted — all anchored on time-to-power rather than $/kWh.
This guide covers where fuel cells beat simple-cycle gas turbines and diesel gensets on schedule, where they lose on capex, what the vendor market actually looks like, and the decision framework that determines when a modular block belongs on your campus single-line.
Can Fuel Cells Actually Power a Data Center? What the 2024–2026 Deal Flow Shows#
The answer is no longer hypothetical. Bloom reports more than 400 MW of SOFC systems deployed at data centers worldwide as of 2024–2025 — a company figure, not an audited one — and three contracts signed since late 2024 confirm procurement at gigawatt scale.
The Bloom–AEP agreement covers up to 1 GW with an initial 100 MW order, co-located at customer sites to serve AI data centers in PJM. Bloom and Equinix have exceeded 100 MW across 19 IBX data centers in six states, with roughly 75 MW operational and 30 MW under construction. The Bloom–Oracle master agreement expanded in April 2026 to up to 2.8 GW, with an initial 1.2 GW contracted.
Note what "contracted" does and does not mean. These are capacity agreements and master frameworks, not energized megawatts. The Equinix number is the most useful of the three precisely because it splits operational from under-construction capacity.
The Quanta Computer Fremont expansion is the clearest signal of motive. After costly utility interconnection delays slowed Quanta's manufacturing expansion, Bloom delivered an islanded, load-following installation that brought the load online without the utility. Oracle has publicly attributed its onsite fuel cell adoption to grid interconnection delays and capacity constraints — the customer saying it, not the vendor.
How Fuel Cells Compare to Gas Turbines and Diesel on Cost, Efficiency, and Lead Time#
Fuel cells lose on sticker price and win on schedule and heat rate. Goldman Sachs benchmarks installed capex at $3,000–$5,000/kW for fuel cells, $800–$1,500/kW for simple-cycle gas turbines, and $300–$700/kW for large diesel standby gensets. On capex alone, fuel cells look uncompetitive against an incumbent turbine or diesel pad.
2x to 6x — the installed capex premium a fuel cell carries over a simple-cycle gas turbine across the Goldman Sachs bands, and roughly 3.5x at the band midpoints. Everything else in the business case has to clear that gap.
The math changes when you net out fuel cost, schedule risk, and emissions permitting. SOFC and MCFC platforms reach 50–65% electrical efficiency (LHV) in power-only mode per vendor literature, which Goldman characterizes as 10–30% more efficient than simple-cycle gas turbines. Heat rates are reported at roughly 6,200–7,000 Btu/kWh for fuel cells against 9,000–10,500 Btu/kWh for simple-cycle turbines in distributed applications — a figure that appears in vendor-sponsored trade content and should be confirmed against your own OEM quotes. Goldman estimates fuel savings of roughly $10–$25/MWh versus simple-cycle gas turbines from the efficiency delta.
Those two vendor figures do not fully reconcile, and the gap is instructive. A heat rate of 6,200–7,000 Btu/kWh implies roughly 49–55% electrical efficiency, which sits below the top of the separately quoted 50–65% range. Run the conversion yourself on any bid before accepting either number.
Against a high-capacity-factor baseload profile, the per-MWh fuel delta compounds into a line item that has to clear the capex premium over the asset life. The case is delivered $/MWh, not day-one installed $/kW.
Comparison Matrix: Fuel Cells vs. Simple-Cycle Gas Turbine vs. Diesel Genset#
| Criterion | Fuel cell (SOFC/MCFC) | Simple-cycle gas turbine | Diesel genset |
|---|---|---|---|
| Installed capex | $3,000–$5,000/kW | $800–$1,500/kW | $300–$700/kW |
| Electrical efficiency, LHV | 50–65%, vendor-stated | 10–30 points below fuel cells | Not in the sourced comparison set |
| Heat rate | ~6,200–7,000 Btu/kWh, sponsored source | 9,000–10,500 Btu/kWh, sponsored source | Not in the sourced comparison set |
| Lead time, order to operation | Under 12 months; Bloom claims 55 days on one Oracle site | 5-plus years for new projects | Shortest; off-the-shelf for standby duty |
| Local emissions | Virtually zero NOx, SOx, PM — no combustion flame, vendor-stated | Combustion NOx drives air-permit scope | Combustion NOx and PM drive air-permit scope |
| Primary fuel | Natural gas, biogas, or hydrogen | Natural gas | Diesel or HVO |
| Typical role | Primary or bridge power | Primary or peaking | Standby and emergency |
Bloom states the AEP deployment will produce 34% lower CO2 than displaced marginal PJM generation — a vendor projection against a modeled counterfactual, not a metered result. Benchmark it against current gas turbine and generator options rather than accepting it as settled.
Can Fuel Cells Run on Natural Gas? Fuel Inputs and the Hydrogen Pathway#
Yes, and natural gas is the commercial default today. Bloom's SOFC platform converts natural gas, biogas, or hydrogen without combustion, and FuelCell Energy's carbonate platforms run on natural gas, alternative fuels and biofuels, or hydrogen.
The dependency moves, it does not disappear. Current commercial deployments depend on natural gas infrastructure availability and pressure. If the site lacks a gas lateral sized for continuous MW draw at delivery pressure, the schedule advantage evaporates behind a gas-side interconnection with its own approval cycle. You have traded an electric utility queue for a gas utility queue — sometimes a good trade, never a free one.
The hydrogen pathway is real but constrained. Vertiv positions hydrogen fuel cells as a zero-emission backup option as diesel permitting tightens. The deployed reference point is a Groningen installation providing about 3.5 MW with roughly 4 hours of backup on an 1,800 kg hydrogen storage system. Scaling that storage profile to campus load is where hydrogen logistics and safety compliance at multi-MW scale becomes binding.
The practical sequence against data center backup power alternatives: natural gas for primary or bridge power now, hydrogen-blend capability as a contract option, full hydrogen only where storage footprint and the authority having jurisdiction can support it.
Who Makes Fuel Cells for Data Centers? Vendor Market and Technology Choice#
The public record is heavily concentrated, and that concentration is itself a procurement risk. Bloom dominates the SOFC primary-power segment with more than 400 MW deployed and all three anchor contracts above. FuelCell Energy has introduced a packaged 12.5 MW utility-grade power block for data centers with manufacturing expansion plans. For hydrogen backup, Vertiv is the integrator-side voice.
Technology maps to role:
- SOFC. High electrical efficiency, natural gas default, primary or bridge power, modular blocks aggregating to tens of MW. Best suited to steady baseload duty.
- MCFC. FuelCell Energy's 12.5 MW blocks, natural gas or biofuel, designed to aggregate into campus-scale power.
- PEM. Hydrogen-fueled backup, shorter runtime profile, positioned for diesel displacement rather than prime power.
If you are sourcing primary power behind the meter, the reference list is short and single-vendor. Price that. Available systems across all three chemistries are indexed in the fuel cells category, and the sourcing standard behind those listings is documented in the methodology.
What Is the Downside of Fuel Cells? The Risks Procurement Must Price In#
Four technical risks belong on the term sheet, plus a fifth that is structural rather than engineering.
Capex. Higher upfront capex than gas turbines is a documented risk. The $3,000–$5,000/kW band does not amortize down on its own; the case must be carried by fuel savings, schedule, and permitting.
Stack life. Stack degradation typically requires replacement every 5–10 years. That is a mid-life capital event with no equivalent on a turbine at the same interval, and the service contract — who owns the stack, who pays, how availability is guaranteed during the swap — is the negotiation surface.
Fuel infrastructure. Deployments depend on gas availability and pressure, and fuel cost and availability remain a documented dependency for the economics. Weak gas service is a schedule problem before it is a price problem.
Hydrogen logistics. Safety compliance at multi-MW scale remains unresolved at large continuous-duty scale.
Vendor concentration. Every anchor prime-power contract in the public record belongs to one supplier. That is a commercial exposure — on pricing, on slot availability, and on what happens to your campus standard if that supplier's order book fills.
Power Risk and Mitigation: Fuel Cells for Data Centers#
| Constraint | Why it matters | Equipment affected | Timeline risk | Buyer action |
|---|---|---|---|---|
| Capex premium vs. gas turbine | TCO depends on fuel savings and avoided interconnection cost | SOFC/MCFC modules, balance of plant | Low | Build the case on delivered $/MWh, not installed $/kW |
| Stack replacement every 5–10 years | Mid-life capital event; contract structure decides who pays | Fuel cell stacks | Medium — downtime during swap | Negotiate stack ownership, replacement schedule, availability guarantee |
| Gas infrastructure dependency | No adequate lateral means no schedule advantage | Gas service, regulator station | High | Confirm pressure and capacity before NTP; treat as critical path |
| Fuel cost exposure | $/MWh hinges on gas basis against the grid alternative | Fuel supply contract | Medium | Hedge gas; model dual-fuel scenarios |
| Hydrogen logistics at multi-MW scale | Storage footprint, AHJ permitting, safety compliance | H2 storage, dispensing, fire suppression | High for full hydrogen | Stage hydrogen as a future option; start on natural gas |
| Single-vendor concentration | Public reference list for prime power is one supplier deep | Whole plant | Medium | Qualify a second platform before committing campus standard |
When Fuel Cells Win: A Buyer Decision Framework#
The fuel cell case is not universal. It turns on three conditions: the interconnection queue is binding at your node, the site has gas, and the operator values schedule certainty over lowest installed $/kW.
Buyer Decision Framework: When Fuel Cells Win#
| Decision question | Choose fuel cells when | Choose an alternative when | Supporting evidence |
|---|---|---|---|
| Is the utility service date the binding constraint on revenue? | In-service date slips well beyond planned NTP and the load must ship | The date is firm and inside the deployment window | Turbines face 5-plus-year lead times; fuel cells deploy in under 12 months |
| Does the site have adequate gas service? | Lateral and pressure support continuous MW load | Gas infrastructure is absent or capacity-constrained | Deployments depend on gas availability and pressure |
| Are air-permit constraints binding on combustion? | NOx, SOx, or PM thresholds make turbine or diesel permitting slow or uncertain | Combustion permitting is achievable on schedule | Fuel cells are vendor-stated at virtually zero NOx, SOx, PM |
| Does the load profile reward high electrical efficiency? | 24/7 baseload at high capacity factor | Peaking or low-utilization standby | 10–30% more efficient; $10–$25/MWh fuel savings |
| Is siting space or acoustics constrained? | Footprint and noise must be minimal for close-in siting | Greenfield campus with setback headroom | Bloom states about 65 dBA at 10 feet |
| Can the operator absorb the capex premium? | Schedule and permitting value outweigh sticker price | Capex efficiency is the binding metric | $3,000–$5,000/kW vs. $800–$1,500/kW |
The staged structure is the common one: fuel cells as bridge power against the utility in-service date, with a longer-term grid or turbine backbone behind them and diesel held for hard standby. Whether that fits depends on your node, and the Power Intelligence Tool models the mix against site-specific gas service and in-service-date assumptions.
What Buyers Should Do Before Issuing an RFP#
A fuel cell RFP is not a generator RFP with the words swapped. Five items belong on the checklist.
- Confirm gas service capacity and pressure at the required continuous MW draw. If the lateral cannot support design load at delivery pressure, the schedule advantage is gone.
- Define the stack-replacement structure. Who owns the stack at year five to ten, who pays for labor and parts, and how is availability guaranteed during the swap? The 5–10 year cycle is the largest non-obvious lifecycle cost.
- Specify load-following requirements explicitly. High-temperature platforms are baseload-comfortable; partial-load tracking depends on pairing with supercapacitors or batteries. Make the integration scope part of the bid, not an assumption.
- Lock the fuel pathway. Natural gas today, optional hydrogen-blend capability tomorrow, full hydrogen only where storage and permitting support it. Do not pay a hydrogen-readiness premium without a site-specific pathway.
- Require independent performance verification. Efficiency, emissions, and availability figures in this category are overwhelmingly vendor-supplied. Put measured acceptance criteria in the contract.
FAQ: Fuel Cells for Data Centers#
Can fuel cells be used to power data centers?#
Yes. Bloom reports more than 400 MW of SOFC systems deployed at data centers worldwide. The Bloom–AEP agreement covers up to 1 GW with an initial 100 MW order, Bloom and Equinix exceed 100 MW across 19 IBX data centers, and Bloom–Oracle covers up to 2.8 GW with 1.2 GW contracted. Read those as contracted capacity, not energized megawatts.
What is the downside of fuel cells for data centers?#
Four. Installed capex runs $3,000–$5,000/kW against $800–$1,500/kW for simple-cycle gas turbines. Stack replacement every 5–10 years is a mid-life capital event. Deployments depend on gas infrastructure availability and pressure. And hydrogen logistics at multi-MW scale remain unresolved for zero-emission pathways. A fifth worth naming: the prime-power reference list is one vendor deep.
Who makes fuel cells for data centers?#
Bloom Energy dominates the SOFC primary-power segment with more than 400 MW deployed. FuelCell Energy has launched a packaged 12.5 MW utility-grade MCFC block for data centers. Vertiv positions hydrogen fuel cells as a zero-emission alternative to diesel as diesel permitting tightens.
Can fuel cells run on natural gas?#
Yes, and natural gas is the commercial default for current data center deployments. Bloom's SOFC platform converts natural gas, biogas, or hydrogen without combustion, and FuelCell Energy's carbonate platforms operate on natural gas, alternative fuels and biofuels, or hydrogen. Hydrogen and biogas are typically contract options rather than the default fuel.
How fast can a fuel cell data center deployment go online?#
Modular systems can be deployed in under 12 months against 5-plus years for new gas turbine projects. Bloom says it delivered a fully operational system to Oracle in 55 days — one site, vendor-reported. Treat it as a best case rather than a planning assumption, and note that gas service readiness, not module delivery, is usually the binding constraint.
Where Buyers Go From Here#
If the utility in-service date has slipped and the load needs to ship, model a modular block against your campus single-line and gas service map before the next RFP goes out. Compare fuel cell, gas turbine, and diesel mixes against site-specific in-service-date, gas pressure, and air-permit constraints in the Power Intelligence Tool, then benchmark the capex delta against the cost of stranded AI capacity waiting on the grid. Available systems across all four generation categories are indexed in the equipment marketplace.