BE: The AI Power War Has a Winner -- and the Balance Sheet Math Most Analysts Are Getting Wrong
Bloom Energy posted $1.065B in Q2 revenue -- up 166% YoY. Then NVIDIA-backed Nebius picked its fuel cells for a 300 MW AI data center. Here is the scenario math at $229 a share.
On August 12, Nebius Group -- an AI cloud company backed by NVIDIA -- announced it had selected Bloom Energy's solid oxide fuel cells to power a 300 megawatt data center in Vineland, New Jersey. Bloom's stock jumped 12 to 13% on the day. The stock had already risen more than 400% from its 52-week low.
The headline reads like a momentum story. It isn't. Bloom's Q2 results, reported July 28, describe something rarer: a genuine operational inflection in a capital-intensive business that has been trying to reach this point for two decades.
Revenue in Q2 came in at $1.065 billion -- a 166% increase year-over-year. Product revenue alone hit $935 million, up 215%. Non-GAAP EPS was $0.78 against a consensus estimate of approximately $0.41, a 91% beat. CEO KR Sridhar noted, with evident satisfaction, that it took Bloom 21 years to deliver its first billion-dollar annual year in 2022. It took another few quarters to do it in a single quarter.
The forward guidance reflected that inflection. Full-year 2026 revenue guidance was raised to $3.9 billion to $4.2 billion, from a prior range of $3.4 billion to $3.8 billion. Non-GAAP EPS guidance: $2.55 to $2.85.
Why AI Needs Bloom -- and Why That Is Not Obvious
Data centers are power-hungry. Everyone knows this. What is less widely understood is the specific nature of that power problem: it is not just about megawatts, it is about reliability, speed, and geography.
The American grid is not built for the power density of modern AI infrastructure. Permitting new grid connections in desirable metro markets can take years. The Texas grid regulator briefly paused approvals for new grid-connected data centers earlier this year -- a signal that regulators are increasingly concerned about the speed of demand growth. The gap between the power data centers need and the power the grid can deliver on a reasonable timeline is widening.
Behind-the-meter power -- on-site generation that bypasses the grid connection problem entirely -- is Bloom's core proposition. Its solid oxide fuel cells run on natural gas (or hydrogen, for customers pursuing decarbonization), generate power quietly, emit significantly less than a conventional combustion plant, and can be operational in months rather than years. That timeline advantage is the product.
The Nebius deal illustrates the appeal. The company said the switch to Bloom "significantly enhanced" the project with "no significant impact" on development timeline. Nebius targets 5 gigawatts of contracted power by end of 2026 and described the Bloom partnership as capable of "unlocking and expediting" additional sites. A 300 MW deployment is not a pilot -- it is a meaningful step toward that 5 GW target.
Sridhar's comment on the Q2 call deserves attention: "All the major US hyperscalers and over a dozen US neoclouds, AI labs, and colocation data center operators have validated and approved our power solutions." The list is not public. But "validated and approved" suggests Bloom is operating as a preferred vendor -- not a company pitching its first contracts -- to a significant portion of the AI infrastructure buildout.
The $25 Billion Framework
Six weeks before the Nebius announcement, on June 30, Bloom and Brookfield Asset Management disclosed an expansion of their power supply partnership from $5 billion to $25 billion. Brookfield's dedicated AI infrastructure fund targets $100 billion in total commitments. The fivefold expansion was incorporated into Bloom's Q2 guidance assumptions.
A $25 billion framework is not a firm order book. It is a commitment to prioritize Bloom as the fuel cell provider across Brookfield's AI infrastructure pipeline, which spans North America, Europe, and Asia-Pacific. The distinction matters for modeling purposes. But the directionality is clear: the people deploying the most capital into AI infrastructure have committed to Bloom as a primary power vendor at a scale that would have seemed improbable two years ago.
The $20 Billion Backlog and the Price
Total backlog stands at $20 billion, including approximately $6 billion in product backlog. These are contracted commitments, not expressions of interest.
The stock trades at $229.94, against a 52-week range of $40.56 to $351.28. Market capitalization is approximately $67.7 billion. Analyst consensus, across 29 polled, is Buy with an average price target of approximately $275 -- implying roughly 20% upside from current levels. The highest target is $390. The lowest is $97.
That spread tells you where the analytic argument lives. The scenario math -- where $229 actually sits in the distribution of outcomes -- is what follows.
The rest of this briefing is for paid members: the balance sheet analysis most retail coverage is getting wrong, the real share count and dilution picture, three scenario price zones with underlying math, and the catalyst calendar for the next 90 days.
AlphaBriefing Paid gets you every investment thesis, scenario framework, and catalyst brief we publish -- the analysis private intel clients pay four figures for, at a fraction of that.