Bloom Energy's $25 billion Brookfield shelf meets a $20 billion backlog

Bloom Energy has $25 billion of third-party project financing lined up and a backlog near $20 billion to spend it on, while Plug Power told investors it still hasn't decided whether to sell into data centers at all.

The two fuel-cell makers both sell devices that skip combustion and turn fuel directly into electricity. They are now pointed at opposite ends of the same market. Bloom Energy (NYSE: BE) is building on-site power plants for AI factories; Plug Power (NASDAQ: PLUG) is staying with forklifts, electrolyzers and hydrogen production.

"Bloom is now a standard for AI onsite power," CEO KR Sridhar said in the company's second-quarter 2026 earnings release, adding that every major U.S. hyperscaler and more than a dozen neoclouds, AI labs and colocation operators have validated and approved its power systems.

The financing is the part that changes the math. Brookfield Asset Management expanded its agreement with Bloom in June to fund AI-related power infrastructure, lifting the total commitment to $25 billion — a fivefold increase from the prior shelf. That money is project-level capital, not a corporate loan, which means Bloom can build generation on a customer's site without carrying the full construction cost on its own balance sheet.

Bloom's second quarter showed what the demand looks like when it converts. Revenue reached $1.065 billion, up 166% year over year, with operating income of $240 million and free cash flow of $175 million. Product revenue rose 215.4% from a year earlier. Full-year 2026 guidance was raised to $3.90 billion to $4.20 billion in revenue and $2.55 to $2.85 in non-GAAP earnings per share, with non-GAAP operating income of $800 million to $900 million.

The stock has already repriced for much of that. Shares are up 198.48% year to date and 286.97% over the past 12 months, though they trade about 21% below the 52-week high of $351.28. At a $259.35 quote against forward EPS of $2.73, Bloom trades at roughly 95 times forward earnings — rich for an industrial, defensible only if the growth rate holds. Beta sits at 3.81, making it one of the most volatile large-cap industrials in the market.

Why the grid can't wait for Bloom's customers

The demand side is not a forecast Bloom invented. Data centers consumed about 176 terawatt-hours of U.S. electricity annually as of the most recent estimates, roughly what Texas uses in a year, and credible 2030 projections range from 200 terawatt-hours to more than 1,000. The International Energy Agency puts the increase in data-center energy consumption at 80% between 2020 and 2025. Microsoft, Google, Amazon and Meta spent more than $200 billion on AI infrastructure in 2024 alone, up 62% year over year.

That load is colliding with an interconnection queue that takes years. Bloom's pitch is that a solid-oxide fuel cell installed on-site bypasses substations, transmission lines and the queue entirely — natural gas, biogas or hydrogen in, electricity out, no combustion cycle and no turbine. Sridhar framed the trade-off in compute terms: "When the business case is measured in months of AI compute, the fastest dependable path to power wins." He added that "capacity is not going to be our constraint."

The constraint may be on the other side of the meter. Utilities are filtering speculative requests with million-dollar deposits, and when Texas utility AEP required developers to commit to paying for 85% of requested power, applications fell from 30 gigawatts to 13 gigawatts overnight, according to reporting on the buildout. Projects are also being delayed by local opposition. That cuts both ways for Bloom: fewer projects get built, but the ones that survive pay a premium for speed.

Plug Power's smaller bet, and the capital gap behind it

Plug Power's answer is a different business. Its GenDrive hydrogen fuel cell targets forklifts, pallet jacks and warehouse vehicles — a material-handling base, not a power plant. In July 2026 the company announced a technical collaboration with Microsoft to test its proton exchange membrane fuel cells as a possible replacement for diesel backup generators, but CEO Jose Luis Crespo said on the second-quarter 2026 earnings call that Plug remains focused on material handling, electrolyzers and hydrogen production and has made no decisions on data centers.

Wall Street has priced the divergence. Bloom carries a consensus "Moderate Buy" across 26 analysts with a high target of $354, while Plug Power holds a consensus "Hold" from 20 analysts. The gap is not really about fuel-cell technology — both companies make cells that avoid combustion. It is about who can fund a project pipeline. Bloom's Brookfield shelf lets a hyperscaler sign a power contract without Bloom funding construction from its own cash. Plug Power, still working toward profitability, has no comparable vehicle, which makes a data-center pivot a balance-sheet decision rather than a strategy decision.

Sitting out the AI power cycle is not automatically wrong. Plug's growth story simply runs on hydrogen economics and material-handling replacement cycles, and it will be judged on whether it reaches profitability, not on whether it wins an AI contract. The risk is timing: if data-center power demand keeps compounding at the pace the IEA data shows, the capital and the customer relationships being locked in now will be much harder to win back in 2028.

For Bloom, the next checkpoint is whether the $25 billion shelf turns into disclosed bookings. Three things have to land: 2027 EPS printing above $3 as the roughly $20 billion backlog converts, visible project commitments tied to the Brookfield facility, and hyperscaler validation expanding into the 30 to 40 gigawatts of 2027 AI data-center buildout that analysts are modeling. A pause in hyperscaler capital spending or a change to Inflation Reduction Act credits would compress the multiple quickly at 95 times forward earnings. The consensus target of $276.05 sits barely above the current quote, which tells you the sell side is waiting for the same evidence.

This article is for informational purposes only and does not constitute investment advice.