Here’s what’s on our plate today:

• 🧪 Why the clean-energy IPO boom is really an AI story.

• 📰 2026's worst breaches, AI floods the creator economy, and the AI boom in six charts.

• 🗳️ Poll: Will the clean-energy IPO boom hold?

• 💬 Prompt of the Day: stress-test a clean-energy company's reliance on hyperscaler demand.

Let’s dive in. No floaties needed…

The Laboratory

TL;DR

• Three IPOs, one buyer: FERVO Energy’s $10B+ debut, X-energy’s $1.02B raise, and SOLV Energy’s first solar EPC offering since 2008 share one demand driver: hyperscaler offtake deals that make the projects financeable.

• The aluminum precedent: Like ALCOA absorbing Columbia River dam output in the 1940s, Amazon, Google, Microsoft, and Meta now anchor U.S. power demand. Alphabet bought Intersect Power outright for $4.75B.

• Policy selected for baseload: Trump’s One Big Beautiful Bill Act gutted wind and solar credits but spared geothermal and nuclear, the technologies AI training clusters actually need.

• Households are paying the bill: U.S. residential electricity prices are up by more than 36% since 2020, with wholesale prices near data center clusters rising to 267%.

Why the clean-energy IPO boom is really an AI story

In the 1940s, the Pacific Northwest built the world's largest hydroelectric system, transforming the Columbia River into a power source far greater than the region could immediately consume. At the time, that apparent excess only made sense in the context of another wartime imperative: the United States needed vast quantities of aluminum for aircraft production, and aluminum smelting required enormous, uninterrupted power. This was the time when aluminum companies agreed to absorb whatever the new dams could produce, thereby justifying generation that would otherwise be too large for the region’s immediate needs.

Something similar is now shaping how the U.S. public markets react to power generation plants, and the clearest place to see it is in the past few weeks’ IPO calendar.

On May 13, 2026, FERVO Energy, a Houston-based enhanced-geothermal developer (a company that uses oil-and-gas-style horizontal drilling to extract heat from rock that is not naturally permeable), closed its first day of trading roughly 35% above its issue price for a market cap above $10B, the largest clean-energy IPO in U.S. history. Three weeks earlier, X-energy, a small modular reactor developer (a factory-built nuclear plant substantially smaller than a conventional reactor), raised $1.02B at a price 21% above its marketed range, with Amazon as the lead investor. Two months before that, SOLV Energy, a utility-scale solar and battery storage construction firm, filed for the first U.S. solar engineering-procurement-and-construction (EPC) public offering since 2008, on $1.7B in nine-month revenue and a $6.7B project backlog.

Three different technologies, three different balance sheets, and what appears at first glance to be three different investment theses. However, below the surface, a single variable unifies the growth. That variable is a handful of companies whose electricity demand is growing faster than the grid can supply it.

Four buyers, one demand curve

That handful of buyers is small enough to name. Amazon, Google, Microsoft, and Meta are now the four largest corporate buyers of electricity in the United States, and the scale at which they procure power is increasingly resembling industrial integration rather than utility contracting. In January, Meta announced agreements with Vistra, Oklo, and TerraPower for up to 6.6 GW of nuclear power by 2035, describing itself in the same release as “one of the most significant corporate purchasers of nuclear energy in American history.” A month earlier, Alphabet agreed to acquire clean-energy developer Intersect Power outright for $4.75B in cash plus assumed debt, moving an independent power producer directly onto Google’s parent balance sheet. The relationship before that point had been a partnership; the relationship after it is no longer separable from the buyer’s own infrastructure.

The reason for this kind of integration is straightforward, even if its consequences are not. According to the IEA, global data center electricity consumption is on track to roughly double by 2030, to around 945 TWh, while AI-focused workloads are expected to triple over the same period. The growth is concentrated in the U.S., where, by the end of the decade, data centers will consume more electricity than the country’s aluminum, steel, cement, and chemical industries combined. The smelters of 1944 have a digital descendant, and that descendant is now running out of grid.

The pattern that emerges from those numbers aligns more with the relationship between the Columbia River dams and their aluminum customers than with a standard utility procurement contract. The buyer’s long-term commitment to absorb the output is the precondition for the generator’s existence. FERVO’s $7.2B contracted backlog is the company’s primary asset, with Google holding exclusive rights to its geothermal output until at least March 2028. X-energy’s commercial case rests on offtakes from Amazon, Dow, and Centrica, with the Amazon commitment alone running up to 5 GW by 2039. Remove the hyperscaler demand curve from the prospectuses, and the valuations become difficult to defend in any other terms.

The policy filter that was selected for baseload

This is also where the clean-energy framing starts to mislead. The IPOs were marketed to investors who expected the U.S. government to keep supporting renewable energy at scale: tax credits for solar and wind projects, federal funding for new clean-energy infrastructure, and a broader policy commitment to cutting fossil-fuel use. Much of that support had been built into the Inflation Reduction Act, the 2022 climate law that channeled billions of dollars into clean-energy development. However, the political environment in which the IPOs actually entered is closer to the opposite.

The Trump administration’s One Big Beautiful Bill Act, signed in July 2025, accelerated the phase-out of those same wind and solar tax credits. The technologies that came through the rollback intact were geothermal and nuclear, and their survival had relatively little to do with their climate profiles. Both produce continuous, dispatchable, weather-independent baseload, which is what an industrial smelter would have asked for in 1944, and what an AI training cluster asks for now.

Bloomberg has described geothermal as “the only renewables sector that has escaped the administration’s disdain,” noting that Energy Secretary Chris Wright, formerly CEO of the fracking-services firm Liberty Energy, championed the technology in budget negotiations. Liberty Energy is also an early investor in FERVO.

What the policy environment has selected for, in effect, is what an industrial buyer of continuous baseload would have selected for in the first place. The boom, therefore, qualifies as a clean-energy story only in the narrow technical sense that the generation is low-carbon. The structural logic pricing the IPOs is industrial, and the industrial buyer is AI.

The bills arrive at the household meter

Once a generation is built around a small number of very large buyers, the costs of doing so begin to spread well beyond the companies signing the contracts. According to CNBC, U.S. residential electricity prices have risen more than 36% since 2020, reaching 17.44 cents per kilowatt-hour, and the EIA expects another increase by 2027. Bloomberg's analysis of wholesale prices near data center clusters found jumps of up to 267% over five years. State legislatures in Virginia, New Jersey, and Wisconsin are now drafting rules that would require data centers, rather than ordinary households, to pay for the grid upgrades their demand requires.

This is similar to what the Pacific Northwest went through in the 1950s, once it became clear that cheap public hydroelectric power was effectively subsidizing the aluminum industry. The question of who pays for industrial-scale power, and on what terms, is older than the data center by several decades.

However, what is genuinely new today is the structure of the relationship itself. The hyperscalers are not simply signing long contracts. They are taking equity stakes in their power suppliers and, in some cases, buying them outright, which turns a market-facing power producer into more of an in-house infrastructure unit for the cloud business.

The case for pumping the brakes

A counterargument deserves consideration. The prevailing investor view is that AI’s electricity demand will continue to rise for long enough to justify the multi-decade power contracts signed in 2025 and 2026, although that assumption is not universally shared. Speaking on Constellation Energy’s May 2025 earnings call, CEO Joe Dominguez cautioned against overly optimistic forecasts, warning that he thought that “the load is being overstated” and adding, “We need to pump the brakes here.” His concern reflects a broader uncertainty at the heart of the AI infrastructure boom: whether future demand will ultimately match the scale of investment now being committed.

The IEA's April 2026 update added a related observation: the amount of electricity needed to run a single AI task is falling faster than at almost any comparable point in recent energy history, even as the total amount of AI being run keeps rising. Whether total demand keeps climbing depends on whether cheaper, more efficient AI encourages greater use, as cheaper electricity once did, or whether the technology hits a ceiling before the long contracts come due.

When the infrastructure outlasts the logic

The aluminum precedent is worth holding next to that uncertainty. By the late 1950s, ALCOA was siting its own smelters next to TVA hydroelectric infrastructure on the assumption that aluminum demand would continue to grow without a ceiling. Demand continued to grow for several decades. But when it eventually softened, the smelters were idled, sold, or moved offshore, and the dams kept running, finding new buyers at lower margins. The infrastructure outlasted the industrial logic that built it.

The clean-energy IPO boom has been narrated as a return of public-market appetite for the energy transition. A more precise description is that a generation of generators is being repriced as a derivative of hyperscaler capital expenditure. The IPOs are accurate about demand and an honest expression of how that demand appears from the seller’s perspective. What they leave open is whether the next decade looks more like the durable industrial buildout the prospectuses assume, or more like the slow unwinding the Columbia River eventually went through. The infrastructure tends to persist, leaving open the question of what it ultimately serves.

Bite-Sized Brains

• 2026's worst breaches: A new roundup of the year's biggest hacks and data breaches shows attackers moving faster and hitting bigger targets than ever, with AI sharpening both sides of the fight.

• AI floods the creator economy: AI-generated content is reshaping it, raising hard questions about authenticity, payment, and who actually owns an audience.

• The AI boom in six charts: A new visual breakdown lays out how billions are being spent on AI against still-hypothetical returns, a clear-eyed look at the gap between investment and payoff.

Prompt Of The Day

Tuesday Poll

The Toolkit

• Synthesise: No-code platform for building and monetizing AI-powered digital products without writing a line of code.

• Runway: AI video and image generation tool used by filmmakers and marketers to produce cinematic content from prompts.

• Tabnine: AI coding assistant that runs privately on your stack, useful for teams that can't send code to public AI services.

Rate This Edition