Hyperscalers: Volatility helps the call and hurts the put
Fifty-two years ago, economist and future Nobel laureate Bob Merton published a seminal paper on the valuation of corporate bonds. The Merton model has plenty of shortcomings that the financial literature has spent decades trying to address, but its central intuition is rock-solid: Shareholders are long a call on the company’s assets and bondholders are short a put on those same assets. In other words, equity investment centers on upside potential, while credit is about downside mitigation.
That single insight explains much of what we are seeing recently across the capital structures of hyperscalers – the massive providers of cloud computing and networking resources essential to the AI buildout.
Quarter-to-date, sentiment toward the tech sector has notably improved in the equity market (with the exception of the weakness late last week), buoyed by a decent first-quarter earnings season that calmed investor nerves over the risk of capex overspending. The relief has been visible across the entire AI ecosystem: semiconductors, hyperscalers, and even software, which had been the laggard for much of the prior quarter (see Figure 1).
Figure 1: Sentiment toward the tech sector has notably improved in the equity market so far this quarter …
Figure 2: … but credit investors have shown more restraint
The Merton framework helps explain the bifurcation between equity and credit price moves. A leveraged bet on AI infrastructure with uncertain and potentially volatile payoffs tends to raise the value of the equity call even as it makes the bondholders’ implicit put more risky. In Merton terms, asset values may be increased, but so is the strike price of the put.
For shareholders, the upside justifies the gamble. For bondholders, the downside is real and the upside belongs to someone else. That wedge – the classic asset substitution problem – is what credit investors are increasingly pricing, and until the re-leveraging impulse shows signs of reaching a plateau, the divergence across the capital structure is likely here to stay.
Leveraged loans: The sponsor put is no longer working
Figure 3: Software issuers have weighed on the leveraged loan index
Within the software sector, there has also been clear bifurcation between sponsored and non-sponsored issuers. Although private equity (PE) ownership/sponsorship had been viewed favorably during past periods of weakness, such as the COVID-19 pandemic, this is no longer the case. Loans issued by PE portfolio companies have been materially underperforming those issued by non-sponsored firms. Of course, at the index level one could cite differences in quality and sector composition as the driver of this performance differential, rather than just PE sponsorship. Therefore, to address this critique, we go deeper to show the result in two more robust ways.
First, we conduct a like-for-like comparison by focusing on B rated software loans in the S&P UBS Leveraged Loan Index, constructing one portfolio of sponsored issuers and another of non-sponsored ones. These portfolios control for differences in both credit quality and sector. Figure 4 shows that even within this tightly controlled subset, PE ownership has been a drag on performance.
Figure 4: The underperformance of sponsored issuers holds true even when looking at B rated software loans
Figure 5: A regression-based approach shows significant spread compression between PE-sponsored and non-sponsored loans over the past two years
AAA CLO performance: No spillovers yet from the software sector
While software loans have repriced, the broader leveraged loan market has largely succeeded in isolating this weakness and hasn’t been dragged down in a meaningful way. This containment helps explain the striking resilience of spreads in AAA collateralized loan obligations (CLOs).
So far this year, AAA CLO spreads have outperformed their historical beta to leveraged loans by a wide margin. If the past empirical relationship had held, the 30 basis points (bps) of year-to-date widening in leveraged loan spreads should have pulled AAA CLO discount margins wider by roughly 11 bps. Instead, AAA CLO spreads are actually tighter by 10 bps (see Figure 6).
Figure 6: So far this year, AAA CLOs have outperformed their spread beta to leveraged loans
Aside from the index-level containment, another main reason for the AAA CLO outperformance is structural subordination – that is, the order in which claims are contractually repaid, with senior tranches coming before junior ones. Even where software exposure exists within CLO portfolios, the deep subordination cushion at the AAA level means the bar for losses to reach the senior tranche is exceptionally high. AAA attachment points are deeply out of the money, and CLO portfolios tend to be diversified across hundreds of issuers and dozens of sectors.
Therefore, for AAA CLO spreads to reprice meaningfully wider, the stress would need to extend well beyond a single sector – it would require a systemic credit event producing default losses far above historical norms. That scenario remains firmly outside the baseline, particularly given the continued resilience of U.S. economic growth.
Michael Puempel and Gabriel Cazaubieilh contributed to this report.
