That 14.5% is almost certainly overstating the credit risk given that 100% of a utility’s revenues are not coming straight from data centers, JP Morgan itself admits. But still, the totality of AI risk is going to be closely watched over the coming years.

“The IG corporate debt exposure to AI spending to me is a bit concerning,” one investor said. “If there's a rollover in valuations on the equity side people will start to look at what companies are exposed and what sectors — obviously, it's a lot in tech but there are other sectors to watch too.”

Utilizing utilities

Regulated and unregulated utilities are poised to be one of the biggest indirect AI spenders and the ones involved in data center build outs make up roughly 5.3% of the index today, according to JP Morgan. By comparison, AI-related tech makes up around 7.8%.

US data center electricity consumption is expected to reach between 325-580 Terrawatt Hours (TWh) by 2028 up from roughly 176 TWh currently, according to a report from Lawrence Berkeley National Laboratory. That build out is expected to take 3-4 years for new natural gas turbines and around 10 years for new nuclear facilities, which are preferred for data centers because of its large capacity, carbon free generation and the ability to work around the clock.

“Utilities will be large beneficiaries,” one buysider said. “There are various utilities in good areas and tech companies are going to need power for their AI spending.”

Tech companies are already collaborating with utilities to provide energy for data centers and to restart nuclear programs, which is faster than building from new. Google announced a deal with NextEra Energy, Microsoft tapped Constellation Energy, Amazon has a deal with Talen Energy that was expanded earlier this year and Meta is contracted with Vistra and Constellation.

Entergy has been the most active of those hyperscaler-connected utility issuers this year raising $3.45bn across six tranches this year, including a $1.3bn two-part subordinated deal that priced last week, according to 9fin data.

Others such as Duke Energy and Xcel Energy are reportedly in talks to tap private credit markets for a portion of the financing needs of their $87bn and $60bn capital plans, respectively.

This week alone three electric and gas issuers priced $3.15bn. That includes a $750m tap from Plains AM Pipeline and a $1.65bn three-part tap from Enterprise Products, both of which will be used in part for capex spending to keep up with AI energy demands.

Deals this week have pushed utility issuance overall up to $162.5bn year-to-date as of 12 November, which slightly surpasses full year 2024 supply. Last year, the category remained active well into December pricing nearly another $10bn from November to the end of the year and sources say the sector will continue to be active this year.

“The utility space is funding a lot of the potential AI investment that's necessary,” one banker said. “It’s always one of the most active sectors, and I would expect them to continue to be active in the fourth quarter.”

Junior subordinated debt has been a popular way for utilities to raise funds this year because it receives 50% equity credit, which helps keep leverage low for the borrower while they spend big on capex. Investors are happy to buy the relatively safe debt at higher yields, as 9fin previously reported.

Because tech companies are agreeing to multi-year contracts with these utilities to provide the power, some investors view this space as even safer than investing in the hyperscalers themselves.

“I'd almost rather be in the utilities than the likes of Meta, if you want to play the AI trade,” one portfolio manager said. “The tech capex send is ultimately driven by how good their AI models are. That’s not something your traditional IG investor can call, but I know the utilities have many gigawatts of energy contracted out and if it's not one of the tech companies using it, there'll be someone else out there that can take the energy supply.”

Constructing an AI future

There’s another cohort of AI-effected industrials, telecoms and REITs that make up just around 1% of the index that could also benefit from all this data center spending.

Construction equipment providers such as Caterpillar and Deere could see a boost from the construction of data centers, especially if they require infrastructure changes to the surrounding area.

Much of that additional business will be front weighted in the AI spending cycle where companies with exposure to AI already trade at a premium to their competitors.

For example, Eaton and Regal Rexnord provide various components for data centers and new energy production sites such as switchboards, transformers and electric motors that are expected to see a boost in the early stages when the data centers a being set up. Carrier Global similarly provides commercial heating and ventilation that will be used in these data center facilities.

Those names tend to trade tight to comparable manufacturing names that don’t have AI exposure.

Digital infrastructure REITS such as Equinix and American Tower provide a different kind of exposure to the AI spend, where they build the data centers and then rent them out to their customers.

Equinix (Baa2/BBB+/BBB), for example, plans to increase its leverage by 1x to 4.5x by issuing some $8bn of new debt to fund an expansion of its data center build out, which comes on top of another $8bn of refinancing needs between now and 2030, according to company earnings. The company kicked off those efforts last week with $1.25bn SUNs due 2030 that priced at 85bps over Treasuries.

Anticipating some of these AI needs, many of these issuers may have front loaded the spending. The industrial companies that JP Morgan identified as having AI exposure collectively issued $20bn of debt in both 2023 and 2024, but that figure has dropped to around $10bn year to date, according to 9fin data.