Power Where the Grid Can't ReachThin moat

The specific problem Bloom solves in the AI boom is the grid bottleneck: the electric grid cannot add capacity fast enough to serve the explosion of AI data centers, so the power must come from somewhere else, on-site, now. Connecting a large new data center to the grid — and building the generation and transmission to feed it — can take many years, an eternity in an AI race measured in months. Bloom offers a way around the bottleneck: on-site Energy Servers that generate power at the data center itself, running on piped-in natural gas, independent of the grid's queue and constraints. For a project whose entire timeline is gated by power availability, this is enormously valuable.

This is Bloom's real and genuine strength in the current moment, and it is why hyperscalers and cloud builders are buying its systems in such quantity: it turns a multi-year power-procurement problem into a multi-month one, unblocking billions of dollars of AI investment. The value is real and the demand it generates is real. The reservation is that the grid bottleneck is a temporary condition, not a permanent one: utilities, regulators, and power developers are responding to the same crisis with enormous investment in new capacity, faster interconnection, and behind-the-meter generation, and over the coming years the grid will, in many places, catch up. Bloom's advantage of supplying power where and when the grid cannot is powerful precisely because the grid is currently overwhelmed — which means it is an advantage tied to a bottleneck that the whole energy industry is racing to relieve. It is a genuine strength and a real source of demand today, but one whose durability depends on the grid staying behind, which it will not do forever — even as the wait fills a ~$20B backlog today¹.