IdeaHouse 2026 Part IV – AI’s Power Crisis and the Energy & Infrastructure Plays

By far the most consistent macro pitch at IdeaHouse Omaha 2026 was AI’s power problem. Multiple presenters cited the same arithmetic: a single ultra-large AI data center can require the equivalent of 1-4 million two-bedroom apartments’ worth of power, the United States faces a 36-60 GW shortfall through 2029, and traditional grid generation has a 5–7 year lead time the hyperscalers do not have. Seven IdeaHouse 2026 pitches played the theme directly – from solid-oxide fuel cells that ship in 90 days, to Vaca Muerta oil production at $40/barrel cost, to Canadian heavy oil filling Permian’s plateau, to a hypersonic test platform with a $2.7B annual government addressable market, to a timber REIT quietly pivoting to data-center land sales, to AI-grade NAND.

Daniel Chee Makes the Case for Bloom Energy as AI’s Power Crisis Plays Out

The Power Crisis Behind the Pitch

Chee opened with the demand side of the equation. AI, he argued, is not merely a technology trend; it is a physical infrastructure emergency. A single ultra-large AI data center can require the power equivalent of one to four million two-bedroom apartments. Even mid-scale facilities housing between 500 and 2,000 GPUs still require meaningful dedicated generation capacity. The cumulative effect is a US power shortfall Chee estimated at 36 to 60 gigawatts through 2029.

The conventional response to power shortfalls involves new grid infrastructure: transmission lines, substations, gas peaker plants, and utility-scale renewables. But Chee pointed out that lead times for these solutions run five to seven years. Hyperscalers cannot wait that long. He cited Microsoft, Nvidia, and Google as companies that have each publicly identified power availability as their single biggest operational constraint, ahead of chip supply or software talent.

Why Bloom Energy

Bloom Energy was founded in 2001 and is headquartered in San Jose, California. The company designs and manufactures solid-oxide fuel cells (SOFCs), which generate electricity on-site directly from natural gas, biogas, or hydrogen, without combustion. By 2025, Bloom had installed approximately 1.4 gigawatts of capacity across more than 1,000 locations in nine countries.

The core competitive advantage Chee stressed is speed of deployment. Where traditional power generation requires years of permitting, construction, and grid interconnection, Bloom’s fuel cell systems can be operational in as little as 90 days. The technology is modular, meaning a customer can start with a smaller installation and expand as load grows. That suits the phased build-out schedules of large data center campuses, which rarely go from zero to full load overnight.

Chee also noted that the fuel cells produce virtually no nitrogen oxide or sulfur oxide emissions. The operating chemistry of a solid-oxide cell occurs at temperatures too low for the NOx reaction, and fuel is desulfurized before it enters the stack. For data center operators facing tighter environmental regulations while trying to move fast, that combination of speed and clean output is hard to replicate with a gas turbine or diesel generator.

A Business Growing Into Its Backlog

Chee acknowledged that Bloom Energy is not yet consistently profitable on a GAAP basis, but argued that margin trajectory tells the more important story. Full-year 2025 revenue reached $2.02 billion, up 37 percent from 2024, driven heavily by the AI data center buildout.

Gross margin for the full year came in at 29.0 percent, up from 27.5 percent in 2024, with non-GAAP gross margin at 30.3 percent. Third-quarter 2025 results showed the pace of improvement most clearly: revenue of $519 million represented 57 percent year-over-year growth, while gross margin expanded by more than five percentage points from the prior-year period.

The forward indicator Chee found most compelling is the order backlog. Bloom exited 2025 with a total backlog of approximately $20 billion, split between roughly $6 billion in product sales and the remainder in long-term service contracts. Management has guided for 2026 revenue of $3.1 billion to $3.3 billion. The Oracle partnership sits at the center of that growth:

Oracle contracted for approximately 1.2 gigawatts of Bloom capacity for its AI data center infrastructure, then expanded the relationship to procure up to an additional 2.8 gigawatts, with a 2.45-gigawatt deployment announced for its Project Jupiter site in Dona Ana County, New Mexico. That single deployment alone exceeds Bloom’s entire installed base from a few years ago.

The Hyperscaler Bottleneck and Bloom’s Timing Advantage

Chee returned throughout his pitch to the timing mismatch at the center of his thesis. Hyperscalers are spending hundreds of billions of dollars on AI infrastructure, and the one variable they cannot buy their way out of quickly is power. Traditional grid solutions require permitting, environmental review, interconnection queues, and multi-year construction cycles. Bloom sidesteps most of that. Its systems are deployed on the customer’s property, eliminating grid interconnection as a gating factor entirely. The systems achieve up to five-nines reliability (99.999 percent availability), which compares favorably with the grid itself in many US markets.