The race to supply electricity to AI has been portrayed as a once-in-a-generation infrastructure boom. But recent comments from industry executives suggests the narrative is already evolving. Across utilities, insurance brokers and risk advisers, executives are spending less time talking about simply powering AI and more time explaining who will bear the financial risks that come with building it.

Recent comments from Duke Energy, PG&E, PSEG, Aon and Arthur J. Gallagher all described different pieces of the same transformation.

Rather than simply enabling AI-driven growth, companies are redesigning contracts, financing structures and insurance programs to allocate construction, utilization, liability and capital risks before billions of dollars are committed—or before the first shovel goes into the ground.

For utilities, that starts with ensuring hyperscale customers—not existing ratepayers—bear the incremental costs of AI infrastructure.

Duke Energy executives disclosed this week that the company has secured 7.8 gigawatts of electric service agreements (ESAs) with data center customers and expects another 15.4 gigawatts of projects to move into contract over the next year.

But management repeatedly emphasized that adding new load only works if financial risk is contractually allocated.

"Our contracts ensure large users of energy pay the cost of serving their facilities," CEO Harry Sideris said while outlining Duke's Customer Protection Plus framework.

Chief Financial Officer Brian Savoy added that the agreements include minimum-take provisions designed to protect future revenues.

"Our contracts contain minimum take provisions... protecting existing customers while ensuring the growth ahead provides shared benefits for all."

PG&E described a remarkably similar approach.

While the company expanded its data center pipeline to more than 12 gigawatts, CEO Patti Poppe said management is becoming increasingly selective.

"We're focusing not on size, but on quality."

Projects now require signed work performance agreements and meaningful upfront financial commitments before advancing through the development pipeline.

At the same time, PG&E made clear that AI demand must reduce—not increase—costs for existing customers.

"We're very focused on pricing this load correctly... attractive to data center customers, but still rate reducing for our other customers."

The company also tied its broader capital strategy to California's wildfire liability reforms, arguing that attracting low-cost capital depends on creating a "durable, financeable" liability framework. Without one, management warned it would reevaluate capital allocation priorities rather than continue investing under an uncertain risk regime.

The same logic is increasingly extending beyond electricity into insurance and capital markets.

Aon said last week that one of the world's largest technology companies sought its help because traditional insurance structures were no longer sufficient for AI infrastructure.

"As the client accelerated its investment in large-scale digital infrastructure, traditional risk solutions were no longer sufficient."

Instead, Aon said it combined commercial insurance and reinsurance expertise to "redesign the client's risk financing strategy, expand available capacity and improve operational efficiency."

Executives described this as part of a broader trend, saying technology companies increasingly need partners capable of solving "complex risk, resilience and capital challenges through coordinated solutions." The company also expanded its data center lifecycle insurance program to $5 billion of capacity, supporting digital infrastructure from construction through operation.

Arthur J. Gallagher identified a similar shift, although from the perspective of advisory services rather than assuming additional balance sheet risk.

Rather than emphasizing premium growth, CEO J. Patrick Gallagher Jr. argued that AI infrastructure is creating sustained demand for expertise in structuring increasingly complex risks.

"AI-related infrastructure, including data centers... creates a multiyear opportunity because clients and carriers need expertise, structure and speed as well as market access."

He added that "as risk becomes more complex, the value of Gallagher's expertise becomes more important, not less."

PSEG reached the same conclusion from the financing side of the equation.

Discussing future nuclear development, management argued that long-term investment will depend on "an appropriate allocation of project risk," reinforcing the view that financing AI infrastructure increasingly depends on determining where project risk ultimately resides before construction begins.

THIS WEEK'S RMN

MARKETS · Spokane Wildfires Reveal How Costs Are Becoming a Utility Financing Question
Analysts quickly shifted from wildfire damage to cost recovery, securitization and insurance, highlighting how investors increasingly view wildfires as a utility financing challenge rather than simply an operational risk. — Read →

MARKETS · KKR Sees Lower Insurance Returns, Shifts Gears to Focus On Alternatives
KKR says it is slowing insurance capital deployment while betting that proprietary asset origination—not insurance balance sheet size—will determine the next winners — Read →

Risky Science

Insurance Isn't Just Transferring Risk Anymore

Most discussions about private credit and life insurance focus on ownership. Why are alternative asset managers buying insurers?

How much private credit belongs on insurer balance sheets? Are current valuations sustainable?

In the latest episode of Risky Science, we spoke with Andrew Granato, Assistant Professor of Law at the University of Texas, and Pranjal Yadav, a finance researcher at Yale, about why they believe those may be the wrong questions.

Their research argues that life insurance has become more than a traditional risk-transfer business. Increasingly, insurance regulation, guaranty funds and insurer balance sheets form part of the financial infrastructure supporting private credit markets. As Granato explains, the prevailing narrative has focused on insurers as a source of "permanent capital," but his research asks readers to look deeper.

"The point of our article is not necessarily to dispute that narrative, but to qualify it," Granato said. "The structure of life insurance insolvency, tax and financial regulation really amplifies the incentives for private equity firms to pair insurers with opaque private credit and ratchet up the risk on life insurer balance sheets."

The conversation also makes clear that the researchers are not arguing private credit has no place in insurance portfolios.

"It would be a mistake to suggest that private credit does not belong on the insurer balance sheet," Yadav said. "Private credit is a big asset class... insurers are in a good position to capture that premium."

Instead, they argue the regulatory framework has not evolved as quickly as the market itself.

One of the most striking moments comes when Granato discusses what he sees as an overlooked financial stability issue.

"There's a sheer lack of precedent for managing a large-scale insurer insolvency through the guaranty fund mechanism," he said. "AIG would have been the first time that happened, but TARP forestalled that from having to occur."

That shifts the discussion beyond private credit performance and toward systemic preparedness. Previous insurer failures involved companies measured in billions of dollars. Today's largest life insurers manage hundreds of billions in assets, yet the guaranty fund system has never been tested against a failure on that scale.

The discusison also explores private credit valuation, private ratings, shadow reinsurance, risk-based capital rules and why current supervisory tools may struggle to keep pace with increasingly complex insurer balance sheets.

For investors, the broader implication is that insurance is no longer simply another financial sector. It has become an increasingly important transmission mechanism between private credit, capital markets and financial stability.

Whether current regulation has kept pace with that transformation may become one of the defining questions of the next credit cycle.