Chips Were the Trade, Electrons Are the Bill: What the Energy IPO Rush Opens Up

What Happened

US energy companies raised $12.6 billion through IPOs in the first half of 2026. All of 2025 produced $4.3 billion, so six months brought in nearly three times a full prior year, and it stands as the largest first half since 1999 (OilPrice, Crypto Briefing).

The composition of that list is the interesting part. The largest deal was geothermal developer Fervo Energy, which upsized its Nasdaq offering from 55.5 million to 70 million shares and priced at $27, raising $1.89 billion on May 13 with J.P. Morgan, BofA Securities, RBC Capital Markets and Barclays leading (Fervo Energy). Dealogic figures put electrical equipment maker Forgent Power Solutions at $1.51 billion and solar contractor SOLV Energy at $512 million (Crypto Briefing). This is not the old oil-and-gas listing cycle riding a commodity price. It is geothermal, nuclear and construction, companies whose job is to produce electrons and push them into a data center.

The demand math explains the crowd. A single AI-focused data center consumes roughly 876,000 megawatt-hours a year, about what every household in Salt Lake City uses. US electricity consumption is forecast to rise 39% between now and 2035 (Finimize). RBC clean energy analyst Chris Dendrinos framed the shift simply: once investors worked out how much power the chips need, a huge tailwind formed behind these companies (OilPrice).

Post-listing performance tells a different story. Close to two-thirds of the energy companies that went public this year and last now trade below their IPO price, against 40% across all sectors (Finimize). The money has crossed over to power. What that money should pay for it is still unsettled.

What This Means for Founders

The takeaway is not “go build a power plant.” What those underwater listings reveal is that public markets have stopped paying for capacity plans and started paying for delivered electrons. Geothermal drilling, reactor licensing, transformer lead times: every one of those runs on a multi-year clock, and no venture-scale team should try to ride it. The gap that opens during those years is the actual market.

Look at where projects die. More than $130 billion of AI data center projects were delayed or blocked in the first quarter of 2026 alone (Crypto Briefing). Capital was rarely the reason. Interconnection queues, permitting, water rights, transformer availability and local opposition were. In PJM and ERCOT territory the wait for a grid connection is measured in years, which is why hyperscalers keep signing behind-the-meter deals and why the phrase “power available in 2029” now shows up in site brochures the way square footage used to.

That is a services and software gap, not a capital gap. Interconnection queue tracking, procurement diligence, load-shifting schedulers, waste heat recovery, site selection data, contract templates for demand response. None of it gets built in-house by a developer preparing an S-1, because the work is margin-thin and headcount-heavy, exactly the profile that gets outsourced. The companies raising billions this year are, functionally, a customer list that just published its constraints in a public filing.

What You Can Do Now

Read the S-1s as customer research rather than investment research. The risk factors section names precisely what is slowing each company down, and each of those sentences is a product requirement someone will pay to remove. If you sell anything adjacent to data centers, start with the interconnection queue position of your target sites, not the land price, because queue order sets the schedule. And if you sell hardware or inference, break power out as its own line in your unit economics now. It was buried in overhead last quarter. It shows up as a separate negotiation in the next contract.