Bloom Energy Slides as Chevron Gas and DOE Nuclear Crash Its AI-Power Trade

The Two Headlines That Did It

On June 23, Bloom Energy fell as much as 13% from the prior session's record close near $345.85, dragging the HIP-3 perp down 13.81% to roughly $297.30. Nothing broke at Bloom — the news was about everyone else.

First, Microsoft and Chevron unveiled a 20-year power agreement to build "Project Kilby," an estimated $7 billion, 2.67 GW natural-gas plant near Pecos, Texas, co-located with a Microsoft AI data center and built on GE Vernova and Caterpillar turbines, with first power targeted for 2028. The same morning, the Department of Energy committed up to $17.5 billion in loans to get ten large nuclear reactors under construction by 2030.

The translation traders made instantly: the two biggest buyers of firm power for AI — hyperscalers and the federal government — both just signaled they intend to source it from gas and nuclear, not only fuel cells.

Priced for a Monopoly It Never Had

This is a valuation reckoning, not an earnings miss. Bloom had run up roughly 275% year to date on a clean story: AI data centers need gigawatts now, grid interconnection queues can't deliver, and Bloom's on-site fuel cells fill the gap. That thesis is still intact — but the tape had quietly priced Bloom as *the* answer rather than *an* answer.

When Microsoft picked gas and Washington bankrolled nuclear on the same day, the market was forced to acknowledge the AI-power pie has many claimants. After a move that size into an all-time high, you don't need a company-specific problem to trigger profit-taking; you just need a reason to question the monopoly premium.

This is also the second such scare this month. Shares already dropped about 10% on June 10 when the Cheyenne data-center project changed course, stoking fears that AI capex itself could slow. The stock is increasingly trading on the durability of a narrative, not the next print.

The Narrative Cracked, Not the Business

Separate the tape from the fundamentals. Bloom's actual results have been strong — Q1 2026 delivered record revenue of $751 million, up 130% year over year, an EPS beat, and raised full-year guidance, backed by a backlog north of $20 billion from Oracle, Nebius, and Brookfield deals. None of that reversed on June 23.

What changed is the ceiling. If gas turbines and nuclear capture a meaningful share of hyperscaler load, Bloom's addressable slice of the AI build-out is large but no longer unbounded — and a stock carrying a triple-digit forward multiple has almost no room for "large but not unlimited." The read-throughs that matter next are Chevron's final investment decision on Kilby, expected by the end of 2026, and whether any other hyperscaler follows Microsoft toward behind-the-meter gas.

On the HIP-3 perp, the move lands harder than the headline suggests: with 24h volume near $1.1 million, the book is thin, so a re-rating like this gets absorbed in sharp, oracle-driven repricings rather than orderly drift — and it keeps ticking through nights and weekends while the underlying NYSE shares are closed.