Key Takeaways:
Credit heavyweights are loading up on debt designed to survive a deep cycle, preparing for what they see as an inevitable AI credit bust.
DoubleLine Capital and Oaktree Capital Management are buying bonds tied to artificial intelligence infrastructure that can hold up if the technology boom turns into a credit bust, even as they warn the probability of a speculative bubble approaches certainty.
"The probability that we will be in an AI bubble? I'll put maybe 100% on that," Robert Cohen, portfolio manager at DoubleLine, said Wednesday at the Bloomberg Global Credit Forum in New York. "You have to think about what credit will survive a deep cycle. You want credits that either through structure or just a very strong balance sheet will survive."
The caution comes as a tidal wave of debt floods the market. US hyperscalers — the big technology firms driving AI buildout — have sold more than $155 billion of unsecured bonds globally this year, already up more than 45% from their entire 2025 issuance, according to a May 21 report by Barclays. Bloomberg Intelligence estimates companies will spend roughly $5 trillion on AI capital expenditure over the next five years, much of it financed through borrowing.
The scale of financing is testing the boundaries of credit markets. This week alone, Hut 8 Corp., a data center operator, sold about $4 billion of investment-grade bonds for a Texas project linked to Nvidia, drawing $17 billion of investor orders — four times the offering size. A separate $36 billion bond sale to fund chip procurement for Anthropic, the AI model developer backed by Amazon, is nearing completion. Apollo Global Management and Blackstone Inc. recently finalized a $35 billion financing package for Anthropic to expand its AI infrastructure.
'Early Innings' of a Financing Boom
Oaktree is investing as if speculative excess may already be building, even if it cannot yet confirm one exists. The market remains in the early stages of data center financing, said Christina Lee, co-portfolio manager in private credit at Oaktree.
"We need to be selective because we really don't know yet who the winners and losers of this competitive set will be," Lee said. "Data center financing is a really large and growing opportunity set. We're just in the early innings of it."
Howard Marks, Oaktree's co-founder, warned in a December note that investors face a binary risk. "No one should go all-in without acknowledging that they face the risk of ruin if things go badly," Marks wrote. "But by the same token, no one should stay all-out and risk missing out on one of the great technological steps forward."
Pimco and Bridgewater Sound Warnings
Pacific Investment Management Co. is taking a more defensive posture. Dan Ivascyn, group chief investment officer, said in a late May video that losses on defaulted AI debt could exceed what investors are accustomed to. He said Pimco is not overweight the sector but sees opportunities in selective buying.
"It's not a sector where we want to be overweight just given the uncertainty, the volatility, the need to predict how companies are going to make money in this space," Ivascyn said. "But because of the massive funding needs, you can be defensive in terms of overall exposure and unlock tremendous value."
Ray Dalio, founder of Bridgewater Associates, said on Bloomberg Television this week that major technology revolutions have historically produced speculative excess. "Nobody can get it exactly right," Dalio said. "You have to either spend a ton of money to capture your market share and don't worry about whether it's too much or not, or you don't spend enough money and you lose your market share."
Cohen at DoubleLine defines a credit bubble as when investors provide debt financing to companies that need real growth just to service their obligations. He said that dynamic is already visible in parts of the AI supply chain. The last comparable technology-driven credit cycle was the dot-com era, when telecom and fiber-optic companies issued billions in bonds that later defaulted as overcapacity crushed returns.
For asset managers, the challenge is distinguishing between companies with durable business models and those dependent on continuous capital markets access. Data centers face a heightened risk of overbuilding because the facilities take years to construct and dozens of projects are launching simultaneously, Cohen said. The bonds being sold today often have maturities stretching decades — far beyond the current technology cycle.
The flood of supply is creating opportunities for the largest managers. Pimco is hoping to leverage its size to capture compelling deals, Ivascyn said. Oaktree is building positions selectively. The question for the broader market is whether the $155 billion of hyperscaler debt issued this year — and the trillions more expected — will be absorbed without a repricing of risk.
This article is for informational purposes only and does not constitute investment advice.
