Carlyle Warns Private Credit’s AI Push Raises Concentration Risk

Private credit firms rushing to finance the artificial intelligence buildout risk encountering the same concentration problems that have plagued lenders exposed to software companies, according to Carlyle Group Inc.

The industry may need to provide roughly $1 trillion to finance AI computing infrastructure, according to a Carlyle white paper published Thursday. That’s equivalent to more than half of private credit assets currently under management.

Failing to put clear limits on concentration in AI compute could prove to be “the biggest mistake of all,” the white paper said.

“We’re in a period where the revenue model to date is uncertain,” Mark Jenkins, Carlyle’s co-president and head of global credit and insurance, said in an interview with Bloomberg. “In such an environment, it’s really hard for us as credit investors to say, ‘well, we’re all in.’”

Private credit managers are increasingly being called on to help finance the massive expansion of AI infrastructure, with estimates of the related capital expenditure expected to top $5 trillion through 2030. Funding is taking a range of forms, including data-center construction and power financing, loans backed by the chips powering the technology, and lending to special-purpose vehicles.

Unlike software, credit risk related to data centers and other AI-related assets is more speculative and more likely to be correlated with the broader economy, while many of the financing structures being used remain largely untested, according to the white paper.

For Carlyle, this doesn’t mean staying away from AI investment. “We want to take the risk, but we want to do it in a balanced manner,” Jenkins said.

One of the biggest challenges for lenders is that it remains unclear where the eventual profits from AI will accrue, whether among chipmakers, data centers or companies building applications, he said.

Software experienced a similar boom between 2020 and 2022, accounting for about half of private equity deals over that period, according to the white paper. Lenders poured money into software companies in part because their recurring subscription revenue was viewed as stable and relatively insulated from economic downturns.

Read More: Private Credit’s Software Bet Is Even Bigger Than It Appears

But the rise of generative AI challenged that assumption by exposing software companies to a shared threat of technological obsolescence. Software loans have since struggled in the syndicated market, with borrowers facing difficulty refinancing debt, while some private credit funds were hit with increased redemption requests.

Jenkins sees a similar lesson for the AI buildout: financings that appear diversified can ultimately depend on a relatively small group of companies, with seven or eight high-quality names accounting for the vast majority of the underlying financings he’s seeing.

Understanding the ultimate counterparty, the contracts underpinning a financing and the value of the underlying assets is particularly important, he said.

“People need to be very, very thoughtful as an investor as to what your counterparty exposure is, what that contract says, and what that ultimate asset value is,” Jenkins said. “In a crisis scenario, all of those things are going to really, really matter. They don’t matter when everything’s going well.”