Bloom Energy vs. FuelCell Energy: Which Data Center Fuel Cell Maker Is the Better Buy Right Now?

Bloom Energy reported Q2 2026 revenue of $1.065 billion, up 166% year over year, with gross margin above 34%, a nearly $20 billion backlog, and raised full-year revenue guidance to $3.4 billion-$3.8 billion. Its shares are up about 225% year to date, reflecting strong AI data-center power demand but also a valuation described as priced for perfection. FuelCell Energy remains a higher-risk turnaround: fiscal Q3 revenue fell 29% and gross loss widened to $24.5 million, although backlog reached $3.6 billion, cash exceeded $737 million after a $300 million equity raise, and management targets positive adjusted EBITDA by fiscal 2027.

Analysis

The investable distinction is not merely backlog size but backlog quality and conversion economics. BE’s advantage should compound if hyperscalers value speed-to-power enough to sign capacity-reservation structures, creating a service and financing annuity beyond initial equipment sales; the key diligence item is whether contracted projects carry enforceable deposits, inflation pass-throughs, and acceptable customer concentration. ORCL and NBIS are relevant read-throughs: any AI-capex financing stress or data-center build delay would first impair project conversion rather than headline demand.

BE is increasingly a power-infrastructure execution story, so its valuation can de-rate sharply on even modest delivery, working-capital, or gross-margin misses. The second-order risk is gas-price and environmental-policy sensitivity: fuel-cell power is attractive versus delayed grid access, but less compelling if grid interconnection queues ease, gas rises materially, or customers prioritize zero-carbon generation rather than dispatchable bridge power. CAT, CMI and GE Vernova are plausible substitution beneficiaries where temporary generation or grid equipment is sufficient, while VRT benefits regardless of the selected power source.

FCEL’s cash runway reduces near-term financing urgency but does not validate its backlog as near-term revenue or establish manufacturing returns. Its first data-center order is an option value, not a repeatable channel until additional orders demonstrate standardized deployment, acceptable gross margins, and customer-funded capacity expansion. Consensus may be over-extrapolating BE’s success to FCEL despite materially different operating proof points; FCEL remains vulnerable to dilution if EBITDA breakeven slips beyond management’s target.

Key Decisions for Investors

• Maintain a 1-3 month relative-value bias: long BE / short FCEL, sized beta-neutral. BE has demonstrated commercial conversion while FCEL remains a capital-intensive turnaround; reassess if FCEL reports two consecutive quarters of data-center order wins and improving product gross margin, or if BE cuts revenue/gross-margin guidance.
• Do not chase BE after momentum-driven upside; use a post-results pullback or evidence of backlog-to-revenue conversion to enter. Target a 6-12 month hold, with risk defined by a material slowdown in product growth, gross margin below the low-30% range, or disclosed customer/project deferrals.
• Add VRT as a less binary AI-power exposure on 6-18 month horizons. Cooling and electrical-distribution demand captures AI build activity whether customers choose BE, utility supply, turbines, or generators; trim if hyperscaler capex guidance weakens broadly, particularly from ORCL.
• Keep FCEL on an event-driven watchlist rather than initiate a core long. Upgrade only after disclosure of contract terms, funded manufacturing expansion, and a credible path toward positive EBITDA; dilution, further gross losses, or backlog cancellations would invalidate the asymmetric-upside thesis.