AI Infrastructure Debt Looks Attractive Before Overinvestment Risk Builds

Steven Tananbaum described the current credit setup as a frustrating one: coupons are attractive, defaults still look low enough to justify owning credit, but total returns have been stuck in the low single digits while equities have participated far more meaningfully in the market rally.

His central point was not that credit is uninvestable. It was that the asset class is priced for a narrow outcome. In January, at Davos, Tananbaum said he had viewed the setup as poor for credit and good for equities. By his account, that has largely played out. In a mid-cycle economy expected to grow at 2% or better, with valuations already stretched, historical credit returns tend to come in below the coupon. That is what he said has happened so far.

The reason is asymmetry. Credit investors are not paid for earnings upside in the way equity investors are. If a company raises numbers, equity markets can reward that surprise. If the company disappoints, credit can be heavily penalized. In Tananbaum’s formulation, credit is already priced as if defaults will remain low, while corporate earnings have more room to disappoint than to create compensating upside.

That leaves credit managers with a practical problem: how to capture returns in an asset class where broad beta is not doing much. Lisa Abramowicz framed the positioning choice as defensive, neutral, or risk-on for the rest of the year. An on-screen Bloomberg Global Credit Forum poll put 42% of respondents at neutral, 31% defensive, and 27% risk-on.

Tananbaum grouped the neutral and defensive responses together — roughly 73% of the audience — and connected that caution to demand for opportunistic credit. The popular strategy, he said, has been a manager with a long playbook: not simply distressed, not simply private credit, but a mandate to look for the best opportunities across the market. In a market where dispersion is limited, he said, positioning in credit is about finding alpha.

Tananbaum’s opportunity set is not organized around a simple risk-on call. He described it as “very situational,” with distress showing up where business models are being questioned.

Software is one expected area of distress, in his telling. The pressure is not just valuation; it is whether the business model that supported prior multiples still holds. Telecom and cable are another area where public equity markets have become more skeptical. He cited Comcast as being at or near a 52-week low, with Charter and Cable One in a similar camp, while some of the debt in the sector had not reflected the same stress in the same way.

That gap between equity and debt pricing is where he sees an interesting relationship. His point was not a categorical conclusion that the debt must fail if the equity struggles. It was that investors should pay close attention when securities in the same business appear to be pricing different versions of the same risk.

Abramowicz characterized this as playing the debt-equity structure in creative ways. Tananbaum accepted the framing, adding that the work also happens within industries. He mentioned Altice, saying a recently threatened transaction looked to him like a negotiating ploy that could bring parties to the table sooner if it worked. He said it involved the same team that “brought you Bausch,” a team he respects, while noting that earlier contemplated transactions at Bausch had not reached the finish line from the equity side.

Healthcare is another out-of-favor area he cited. Tananbaum said healthcare has seemed like “a source of funding for technology stocks,” with money leaving the sector for more favored parts of the market. Within that, he identified Tenet as a company where the valuation appeared reasonable despite some volume issues. He called management “top-notch” and said the credit markets, in his view, would finance the company’s entire market capitalization.

The broader method is consistent with the way Tananbaum described investing from the outset: asking who will buy a security later, why they will buy it, and how the “short movie” of an investment is supposed to end. He used the example of a software company levered four times and growing at 11%. A buyer might regret not buying it when growth was 20%, but if the company was then levered five and a quarter or five and a half times and the market mindset was different, the later setup could still attract capital at a 9% or 10% yield.

The sharpest tension in Tananbaum’s view concerned AI infrastructure financing. Abramowicz noted that while credit managers are frustrated with broad returns, people are pouring trillions of dollars into AI infrastructure, and asked whether that prospect worried or excited him.

Tananbaum’s answer was conditional. The details of the terms matter, but the entities backstopping the financings appear to him to be good credits. If investors are underwriting the next two or three years rather than the next five or ten, he said, they may be overcompensated for the risk.

The problem is supply and behavior. He described the technical backdrop as “terrible,” because the market “doesn’t seem to run out of product.” He pointed to a Google financing as suggesting that companies want to stay in front of the financing need, including in relation to equity. The resulting dynamic, in his words, is not merely similar to an arms race. It is one.

Tananbaum did not claim to know whether the infrastructure spending will be justified. He said he was “somewhat agnostic.” But history, in his view, has a poor record when industries with very high payouts overinvest. He named riverboat gambling and undersea cable as examples. Undersea cable, he said, was “a good thing until it wasn’t.”

That history shapes how he wants to invest in the AI buildout. GoldenTree’s approach, as he described it, is to be deliberate and to seek additional assurances from users that they are committed to the projects being financed. The credit may look well protected for a period, but he does not want to ignore a pattern he has seen in other stretched or over-financed industries: at some point, investors have often been able to get strong protections with below-investment-grade pricing.

Abramowicz asked whether that means waiting until the sector “falls out of bed.” Tananbaum rejected total abstention. If the attractive two- or three-year return is real, that is a lot of income to give up. But he distinguished between taking what is available now and preserving flexibility for a better entry point later. He also emphasized that his view could change quickly as evidence changes.

Abundant supply argues for patience. On AI specifically, he said investors should not assume they will run out of opportunities: “There’s going to be something to do in this space,” particularly if the sector continues to be successful.

On the relative value between public and private securities, Tananbaum said the best value GoldenTree sees is in asset-backed assets, a private asset class. He also said private credit looks better than it has over the past 24 to 36 months, partly because some of the more anxious capital has stepped away.

His wording was qualified. He said “some of the more anxious capital,” including open-ended private credit funds, are out of the market. That leaves funds and other private credit buyers who are more deliberate as the main participants. He also pointed to out-of-favor sectors, which by definition trade at discounts and can become places to find attractive entry points.

But he was careful not to overstate the illiquidity premium. When Abramowicz asked whether investors are getting paid for illiquidity, he called that “an always after-the-fact comment.” It seems better, he said, but not dramatically so. On a scale of 10, he put the current compensation at roughly six or six and a half: above average, but not by much.

His comments on nimbleness were similarly blunt. Investors should assume they cannot be nimble in less liquid instruments. Lower prices bring illiquidity; higher prices bring confidence. That, he said, is how credit markets work. The answer is not to pretend liquidity will be there when needed, but to ask more questions, identify the key variables, and be deliberate.

AI is one case where that discipline matters because the information set can become overwhelming. Tananbaum’s prescription was to focus on the larger issues rather than trying to process everything. In a market with abundant supply, the fear of missing the last available opportunity is misplaced.

Tananbaum identified inflation as “probably the biggest risk in the market.” It has affected rates and, through rates, credit more than equities. He said that has surprised him, but he accepted it as the market reality. He also noted that in the 1970s, equities did better than credit, offering at least one example of a bad inflationary environment where equities outperformed.

He did not offer a clean line in the sand for inflation because the data remain uncertain and because geopolitical developments, particularly war, can change the oil picture. One surprise, in his view, is that oil has been calmer than many expected. He said most people thought that if the market got past Memorial Day, oil would be at $125 to $135. Instead, he described it as being in the mid-$90s.

Abramowicz pressed on the dissonance between warnings from energy executives and the apparent indifference of equity and credit strategists. She referred to Chevron’s Mike Wirth warning that shortages could appear in the United States if conditions continued, and said diesel inventories were at their lowest level since 2003, while market strategists often seemed to discount the issue.

Tananbaum said he processes the question by asking how equities are pricing it. In his view, equities are not pricing the futures curve. Oil services names such as Halliburton have had strong years, but some mid-cap names still look to him as if they are pricing $70 to $75 oil. He added that GoldenTree has a company, more than half-owned by the firm, that watches the oil-services M&A market and has bought several companies over the past 12 months. According to what they see there, levels, pricing, multiples, and expectations have not changed that much.

That tells him market participants remain cynical about how long higher oil prices will last. He said it seems like a better opportunity that oil will be higher for longer than what is in the market. Abramowicz summarized the implication as buying mid-cap oil names. Tananbaum broadened it to include suppliers and to the way oil should be treated as an input cost across portfolios, including sectors such as building materials.

His guess is that oil will be higher for longer.

Asked for a historical analog, Tananbaum resisted a dramatic comparison. His view is that the present market resembles many markets: mid-cycle, stretched, and narrow, with little dispersion except in a minority of securities. He estimated that less than 20% of the market may contain the meaningful opportunities. That may feel unusual to many investors, but he said it is actually common. If one counted historical markets, he guessed this type of environment might represent three-quarters of them, and certainly more than half.

What is different this time is AI expense and the question of who the winners and losers will be. To think through that, Tananbaum compared software and AI-related disruption with the fate of legacy media after the internet. The effects were not uniform. TV moved from nominal growth to less than nominal growth, but still growth. Cable programmers continued to generate above-nominal growth for roughly 15 years and remained healthy businesses. Radio became more marginal around 2008 or 2009 and was in real trouble by the late 2010s. Newspapers, he said, peaked around the period of the Tribune transaction in the mid-2000s and then began to be disintermediated.

The point of the analogy is classification. Not every incumbent industry disrupted by a new technology collapses at the same speed, or in the same way. Some slow but survive. Some remain strong for longer than expected. Some become marginal. Some are disintermediated quickly. Tananbaum’s question for software and other AI-exposed sectors is where each one falls on that spectrum.

That framework also explains his broader credit posture. He is not arguing for a simple retreat from risk. He is arguing that in a stretched market, with AI financing expanding aggressively, inflation still unresolved, and sector disruption uneven, investors need to know what they own, where they sit in the capital structure, who the future buyer is, and what assumptions the price already embeds.