Long-Term Service AgreementsNarrow moat

The mechanism of Bloom's sturdiest moat is the long-term service agreement. When customers install Energy Servers, they contract with Bloom to maintain them over an extended period — keeping them running at guaranteed performance, monitoring them remotely, and replacing the fuel-cell stacks as they wear. These agreements convert a one-time hardware sale into a stream of recurring, contracted revenue that continues for years after the initial installation, and they bind the customer to Bloom for the life of the system. The 10-K gives the terms: the operation and maintenance contracts run from 5 to 20 years, subject to termination for convenience on an annual basis,¹ so the annuity is renewed each year rather than locked in. This recurring, higher-margin service revenue is the closest thing in Bloom's business to the predictable, sticky earnings that characterize a genuine moat.

The value of the service annuity compounds with the installed base: every system Bloom deploys adds to the pool of installations generating service revenue, so the current surge in product sales is simultaneously building a much larger future service business. Over time, a large enough installed base could make service a dominant, stable share of revenue — transforming Bloom from a lumpy hardware seller into a company with a substantial recurring-revenue backbone. The catch, spelled out in the threat, is that Bloom's service economics have historically been unreliable — the cost and timing of stack replacements have sometimes run against the company, turning what should be a high-margin annuity into a margin drag or a source of charges. So the long-term service agreements are a genuine and promising foundation for a real moat, and the growing installed base is enlarging them rapidly. But their quality depends on Bloom servicing the base at consistently good margins, which it has not always managed — making this a moat with real potential that has yet to fully prove its economics — the company's first profitable stretch is quarters old, not years².