Bloom Energy positioned as cheaper alternative AI power play versus NuScale and GE Vernova amid diversification pressure.

Leopold Aschenbrenner's Situational Awareness fund sold roughly 35% of its Bloom Energy (NYSE: BE) position in the first quarter of 2026, cutting its stake from 10.1 million shares to approximately 6.5 million, according to SEC 13F filings. The fund reportedly built the position at an average cost under $90 a share. Bloom traded above $350 at its 2026 peak.

Read that again, because it's the part every article about the "24-year-old genius who called the AI power trade" conveniently skips. While retail investors chase BE after a 1,400% one-year run, the man who made the trade famous has been booking profits. And he's not the only one heading for the exit while the crowd files in.

I spent 28 years trading professionally—as a market maker on the CBOE floor and a trader at the CME—and I've read 13F filings the way most people read box scores. Here's what this one actually says, why it doesn't mean what the doom-scrollers think, and where the money is moving next.

Aschenbrenner is a former OpenAI safety researcher who was let go in 2024. Instead of fading into consulting, he published a 165-page essay called *Situational Awareness* arguing that artificial general intelligence was coming faster than markets believed—and that the binding constraint on AI wouldn't be models or chips. It would be electricity. His projection: AI could consume 100 gigawatts of power by 2030, and the existing grid cannot deliver it.

He then did something essay writers don't do. He raised roughly $225 million and launched a hedge fund to trade his own thesis. That fund now manages billions, and when its quarterly filings drop, stocks move. When the fund's Nebius stake was disclosed in May, the stock surged on the news alone. His signature trade: while Wall Street crowded into chipmakers, Aschenbrenner bought megawatts.

According to the 13F record, trackers put the fund's average cost basis under $90 a share. At Bloom's 2026 high above $351, the original position had roughly quadrupled. At its peak, the remaining stake was worth about $2.2 billion. He sold a third of it—into strength, after a 200%-plus gain on his fund's largest, highest-conviction holding.

Before anyone spins this as "Aschenbrenner turns bearish on Bloom," understand what a floor trader sees in that filing. A 13F is a rearview mirror showing quarter-end positions filed weeks later. He may have bought some back. He may have sold more. Nobody outside the fund knows today's book. What we know for certain is this: a disciplined trader who bought under $90 took massive profits into a parabolic move while keeping millions of shares for the ride. That's not a bearish call. That's risk management—what professionals do reflexively and what retail almost never does. On the floor, we had a saying for traders who never book gains: they're not investors, they're collectors. Collectors eventually donate their collections back to the market.

But the trim isn't happening in a vacuum. Bloom executives and directors have been net sellers of roughly $83 million in stock over the past twelve months, with fresh intent-to-sell filings hitting in recent weeks. The people who know this company best are trimming alongside Aschenbrenner.

The valuation is priced for perfection. BE trades around 30 times sales and north of 100 times forward earnings. Wall Street's own targets tell the story—even after Evercore and UBS raised targets to $350, firms like Clear Street, Roth, and Barclays sit at Hold or Neutral in the $270–$290 zone. The stock has already pulled forward years of flawless execution.

The tape is violent. BE dropped 18% in a single session on June 26. It ripped 11% higher on July 6 when Brookfield expanded its financing framework for Bloom-powered AI projects from $5 billion to $25 billion. One recent session spanned $257 to $307—a 19% intraday range. With a beta over 3, this trades like a leveraged AI index, not a power company.

None of this means Bloom is a bad business. It means the easy money in the stock has been made—by someone else, at prices you will never see again.

I'm not writing Bloom's obituary because the underlying thesis remains sound. Bloom's founder and CEO, KR Sridhar, is a former NASA scientist who worked on Mars missions before spending 25 years building this company for exactly this moment. Electricity is a $5.5 trillion global market. Bloom's revenue last year was about $2 billion. Even after a 1,400% run, the company has captured a rounding error of its addressable market.

The backlog says demand isn't the constraint—capacity is. Bloom sits on roughly $20 billion in orders against manufacturing capacity of about one gigawatt today, headed above two gigawatts by year-end. Sridhar describes AI as putting "a hockey stick on a hockey stick"—an acceleration stacked on top of a digital revolution already underway.

The business model is speed. When a data center operator asks the utility for two gigawatts, they're quoted years. Bloom's fuel cells arrive in months. The company delivered 50-plus megawatts to an Oracle data center in Utah in 55 days against a 90-day contract. Oracle has since expanded to a 2.8-gigawatt master agreement. Bloom was named exclusive power provider for Project Jupiter AI cluster—up to 2.45 gigawatts. Nebius signed for 328 megawatts. Brookfield just put a $25 billion financing framework behind the whole build-out. First-quarter revenue grew 130% year-over-year to $751 million. Full-year guidance was raised to $3.4–$3.8 billion. The company is profitable.

This is a real business hitting escape velocity. The problem isn't the company. The problem is what you're asked to pay for it today—after the informed money already got paid.

In a recent interview, Sridhar named the three bottlenecks constraining Bloom's growth: the customer's construction timeline, permitting, and gas supply. A CEO's bottleneck list is a shopping list.

**The molecule: EQT Corp. (NYSE: EQT).** Bloom's fuel cells run on natural gas. Every gigawatt Bloom deploys creates new, permanent gas demand sitting behind the meter at a data center. EQT is the largest natural gas producer in the United States, sitting on the Appalachian basin that feeds the data center corridor. Every Bloom box that ships is a small, permanent bid under EQT's product.

**The pipe: Williams Companies (NYSE: WMB).** A fuel cell without gas supply is an expensive sculpture. Williams moves roughly a third of America's natural gas, and its Transco system is the superhighway serving the exact mid-Atlantic and Southeast markets where hyperscale data centers are breaking ground. Behind-the-meter doesn't bypass the pipeline—it makes the pipeline the whole ballgame. You collect a toll on the thesis, with a dividend while you wait.

**The shortage: GE Vernova (NYSE: GEV).** Ask yourself why Bloom exists as an AI trade at all. Because gas turbines are sold out for years. GE Vernova's turbine backlog stretches toward the end of the decade with pricing power to match. The setup is heads-you-win, tails-you-win: every data center that can't wait for a turbine calls Bloom; every one that can wait pays GE Vernova whatever it asks. The power shortage feeds both companies.

Three situations, three moves. The same thesis Aschenbrenner built his fund on—AI's power hunger is structural and still underpriced—without paying peak-euphoria multiples for the ticker cable news just discovered.

If you don't own BE: Don't chase it above $300. You'd be buying a beta-3 momentum stock at 30 times sales after the smart money trimmed and insiders sold. June 26's 18% single-day drop is your preview of what one bad headline does here. This stock has proven it will hand you double-digit pullbacks on a regular schedule.