Small Modular Reactors: Always Ten Years AwayNarrow moat

Every discussion of powering AI data centres eventually reaches small modular reactors, and hyperscalers have signed agreements with several developers. If SMRs work at scale they are a better answer than fuel cells on almost every dimension — carbon, fuel cost, output density, and duration.

The record is sobering. Only two SMRs are operating commercially anywhere in the world, and both took longer, cost more and delivered lower capacity factors than expected¹. The regulatory path in the United States remains slow, the supply chain is immature, and first-of-a-kind projects in nuclear have a long history of arriving late.

For Bloom this is a threat with a very long fuse and a real strategic implication. Fuel cells are being bought as a bridge — power now, while the grid catches up or something better arrives. A bridge is a genuine business and it is not a permanent franchise, which is precisely why the installed base and service annuity matter more to the moat than the technology does: those revenues persist after the bridge stops being built.

Watch announced SMR commercial operation dates against their delivery. Each slipped date extends Bloom's window; a first American SMR delivering on time and on budget would be the event that changes how everyone models the second half of this decade.