Pivot Energy AI Briefing — May 10, 2026

Show Notes

Hosts: Rohan Kapoor & Amara Diallo In this episode: • Welcome to Pivot Energy for Sunday, May 10, 2026. I'm Rohan Kapoor. • And I'm Amara Diallo. Today we're talking about how AI is rewriting the rules of the American power grid, from a resurrected nuclear plant to a boardr... • Let's start with the headline number: Three Mile Island, the site of America's worst commercial nuclear accident in 1979, is being brought back online... • It's a remarkable arc. A site that became shorthand for nuclear risk is now being rehabilitated specifically because chatbots and inference clusters n... • The economics are what matter here. Constellation's restart was originally pegged around $1.6 billion, with a 20-year power purchase agreement underwr... Subscribe to the newsletter at pivotnews.ai for the full written briefing.

What is Pivot Energy — AI News Daily?

Daily AI news for energy and sustainability professionals. Two hosts cover how AI is powering the clean energy transition, grid management, and climate solutions.

Rohan Kapoor: Welcome to Pivot Energy for Sunday, May 10, 2026. I'm Rohan Kapoor.

Amara Diallo: And I'm Amara Diallo. Today we're talking about how AI is rewriting the rules of the American power grid, from a resurrected nuclear plant to a boardroom fight that could redefine atomic data centers.

Rohan Kapoor: Let's start with the headline number: Three Mile Island, the site of America's worst commercial nuclear accident in 1979, is being brought back online by mid-2027. The buyer of the output? Hyperscalers running AI workloads.

Amara Diallo: It's a remarkable arc. A site that became shorthand for nuclear risk is now being rehabilitated specifically because chatbots and inference clusters need round-the-clock carbon-free power. Symbolism aside, this tells you how tight the supply situation has gotten.

Rohan Kapoor: The economics are what matter here. Constellation's restart was originally pegged around $1.6 billion, with a 20-year power purchase agreement underwriting it. That's a long-duration contract for roughly 835 megawatts of baseload. For context, that's enough to power one large hyperscale campus, not a fleet.

Amara Diallo: And that's the tell, Rohan. If a single AI customer can absorb the entire output of a reactor, you start to understand why utilities are scrambling. The unit of demand has changed.

Rohan Kapoor: Which brings us to Duke Energy's first quarter. Duke signed 2.7 gigawatts of new data center agreements in Q1 alone, bringing total executed contracts to 7.6 gigawatts. Two-thirds of that pipeline is already under construction.

Amara Diallo: 7.6 gigawatts is roughly nine Three Mile Islands worth of demand sitting in one utility's queue. And Duke serves the Carolinas, which have quietly become one of the hottest data center corridors in the country.

Rohan Kapoor: What's notable is that Duke kept its $103 billion capital plan unchanged. Management is signaling that even with this surge in contracted load, they can absorb it within the existing investment envelope, at least for now.

Amara Diallo: Though that discipline cuts both ways. If demand keeps accelerating, holding capex flat could mean longer interconnection queues and higher prices for everyone else on the system.

Rohan Kapoor: That's the regulatory question. Who pays for the grid upgrades? Duke and its peers are increasingly negotiating special tariffs that put more of the cost burden on the data center customers themselves, rather than residential ratepayers.

Amara Diallo: And those tariff structures are becoming a competitive battleground between states. Virginia, Texas, the Carolinas, Arizona — each is trying to attract AI infrastructure while protecting consumers. The utilities that get this balance right will shape the next decade of industrial policy.

Rohan Kapoor: Speaking of getting the balance right, our third story is messier. Toby Neugebauer, who was ousted as CEO of Fermi Inc., has called a special meeting for May 29 to expand the board and install himself and allied directors.

Amara Diallo: Fermi is one of the more ambitious bets out there — a developer trying to build nuclear-powered data center campuses from scratch. The idea is to co-locate generation and compute behind the meter, bypassing transmission constraints entirely.

Rohan Kapoor: And the proxy fight matters because the strategy is capital intensive and execution-sensitive. Fermi went public last year on the promise of multi-gigawatt sites anchored by small modular reactors and natural gas bridge generation. Investors are now being asked to pick a management team mid-build.

Amara Diallo: What's the read on Neugebauer's case?

Rohan Kapoor: He's arguing the current board has drifted from the founding vision and slowed deal execution. The board, in turn, has cited governance concerns. Without seeing the proxy materials in detail, the thing to watch is whether any of the anchor data center customers weigh in publicly before May 29.

Amara Diallo: Because customer confidence is really the asset here. Nuclear-powered data centers are a bet on a 10 to 15 year build cycle. Hyperscalers signing those offtakes need to believe the company will still be executing the same plan in 2030.

Rohan Kapoor: And there's a broader signal in this fight. A year ago, nuclear-plus-AI was a thesis. Now it's a category with public companies, proxy battles, and competing business models — behind-the-meter developers like Fermi versus utility-scale restarts like Three Mile Island.

Amara Diallo: Two very different theories of how to deliver clean baseload to AI. The restart model is faster and cheaper per megawatt but limited by how many dormant reactors exist. The greenfield model is slower and riskier but potentially unlimited in scale.

Rohan Kapoor: Let's bring it back to what business leaders should take away. First, power availability is now a gating factor for AI strategy, not an afterthought. If you're planning compute expansion in 2027 and beyond, your site selection conversation starts with the utility, not the real estate broker.

Amara Diallo: Second, the financing models are evolving fast. Long-dated PPAs, special industrial tariffs, and equity stakes in generation are all on the table. The companies treating energy procurement as a core competency are pulling ahead.

Rohan Kapoor: And third, watch the governance side. The Fermi situation is a reminder that these are still early-stage businesses executing extraordinarily complex projects. Management risk is real, and proxy outcomes can reset customer and investor expectations overnight.

Amara Diallo: The through-line is that AI's power hunger is forcing the energy sector to move faster than it has in a generation. Old plants are coming back, new ones are being proposed, and the capital is following.

Rohan Kapoor: We'll be tracking the Fermi vote on May 29 and Duke's next quarterly update. That's it for today's Pivot Energy briefing.

Amara Diallo: Thanks for listening. We'll see you next week.