Introduction

Each year, we return to the question of where we are in the credit cycle. This time, the signals point in different directions, and we believe the outlook depends in part on the duration and severity of the war in the Middle East. Inflation, pressure on consumer spending and household savings, and higher energy costs remain concerns, while corporate earnings and balance sheets have, in aggregate, held up well. At the same time, rising sovereign borrowing needs and substantial AI-related capital expenditure are competing for investors’ capital.

For credit markets, the tug-of-war which has characterised recent months persists. Government bond yields have risen, in some cases to the highest levels in more than 20 years,¹ changing the return outlook for fixed-rate bonds, but credit spreads have not widened commensurately. A larger share of higher-quality issuers, healthy cash balances and the refinancing of debt before rates rose have all helped borrowers withstand a more demanding environment. Even so, tight spreads leave less room for disappointment. If growth weakens or financing costs remain elevated for longer, the effects are unlikely to be felt evenly across issuers.

In this paper, we outline why we believe this is a cycle characterised more by dispersion than broad-based credit deterioration. We look across the credit spectrum to assess where resilience persists, where risks are building and where selective opportunities may emerge.

Private credit: beyond the headlines

Turning first to private credit, a series of high-profile defaults in software and software-as-a-service (SaaS) in 2026 has intensified scrutiny of the asset class. These cases have raised questions about its resilience, but the broader market data tells a more nuanced story.

The Cliffwater Direct Lending Index (CDLI), which covers more than 23,000 loans, returned 1.89% in the second quarter of 2026 and 7.71% over the preceding 12 months. Income remained the main driver of returns, while realised credit losses of 0.47% over that period were below the index’s long-run annual average of around 0.98%. At the same time, there are signs that individual deals warrant closer scrutiny. The CDLI’s non-accrual rate rose from 1.27% of cost at the end of 2025 to 1.85% in the first quarter of 2026, before easing to 1.73% in the second. Implied recovery rates fell from around 53% to 49%, and the index recorded unrealised losses of 1.59% over the trailing year.² None of these measures, taken alone, points to broad-based deterioration. Together, though, they reinforce the case for looking beneath the aggregate figures.

The Technology sector illustrates the importance of borrower-level analysis. The sector accounted for 24% of the CDLI at the end of June, and Cliffwater identified AI headwinds as a contributor to second-quarter performance. Crucially, AI’s effects differ across borrowers. Mission-critical enterprise software, deeply integrated into customers’ operations and difficult to replace, is a different credit proposition from a commoditised point-solution SaaS product. Lenders therefore need to assess whether AI adoption supports a borrower’s pricing power and customer retention or poses a threat to both.

The same scrutiny must apply to payment-in-kind (PIK) interest. Allowing a borrower to accrue interest instead of paying it in cash can offer flexibility, but it can also defer a debt-servicing problem. Across the CDLI, PIK income was 6.92% of total income in the second quarter, close to post-COVID lows. While the aggregate figure does not point to widespread increasing reliance on PIK, the important question is whether its use is concentrated among borrowers already under pressure.

We believe the income case for direct lending will potentially remain attractive, but at this stage of the cycle, managers must distinguish borrowers able to withstand technological and economic change from those whose apparent resilience depends on favourable assumptions.

High yield and leveraged loans: tight pricing, different vulnerabilities

In high yield, fundamentals remain reasonably sound in aggregate, yet valuations leave limited protection if they weaken. One measure of this is the spread investors receive for each turn of issuer leverage. This has compressed from the peaks seen after the Global Financial Crisis (GFC) and during the pandemic to among the tightest readings in the available data, as shown in Figure 1.

Figure 1: Spread-to-worst per turn of leverage has significantly compressed

Source: JP Morgan, as of September 2026.

Problems loading this infographic? - Please click here

Meanwhile, net debt relative to earnings has risen modestly from its 2022-23 lows toward approximately 3.9-4.0 times.³ While it remains below the levels reached at previous crisis peaks, investors are receiving less spread for each unit of leverage than they did earlier in the cycle. There is some justification for this as the high-yield bond market has migrated towards a stronger ratings mix in recent years. It now has an average rating of BB and less than 7% in CCC issuers.⁴ Many bond borrowers also locked in low fixed coupons in 2020 and 2021. Their interest costs are likely to rise as those bonds mature and are refinanced, but the adjustment should be staggered.

Leveraged-loan borrowers face a different timetable. Their floating-rate debt passes changes in benchmark rates into interest expense much more quickly. Even without a sharp downturn in revenue, sustained high rates may erode interest coverage and reduce financial flexibility. The broad loan market also has a weaker average rating mix, closer to single-B than the BB average for global high-yield bonds. Those differences help to explain why leveraged-loan borrowers may face cash-flow pressure sooner than fixed-rate high-yield bond issuers, even though loans are generally senior in the capital structure.

More recently, lower-quality segments have begun to reprice, suggesting investors are starting to factor in slower growth and the cumulative effect of higher rates. Even so, the market has remained bifurcated. The ratio of CCC to BB spreads is trading at the 98th percentile of its range over the past 10 years, while higher-quality credit remained expensive.⁵

Figure 2: Lower-quality segments have begun to reprice

Source: Bloomberg, as of September 2026.

Problems loading this infographic? - Please click here

Opportunistic credit and AI infrastructure

One area of opportunity may lie between conventional performing credit and deep distress. Borrowers facing a refinancing gap, a complicated capital structure or a temporary liquidity squeeze may need a tailored financing solution even if the broader market avoids a default wave. More than US$100 billion has reportedly been raised for distressed and opportunistic strategies over the past two years.⁶ That capital may be most useful in a selective, uneven downturn, where underwriting individual situations matters more than making a single call on the whole market.

AI infrastructure financing offers a high-profile test of that discipline. Investment in data centres and computing capacity is creating substantial financing demand, but the risks differ sharply across borrowers. Some neocloud providers combine rapid growth with high leverage and near-term refinancing needs, while the assets securing their debt may have shorter economic lives than the loans themselves. Revenue growth is important, but lenders also need confidence in cash conversion, collateral values and the durability of customer demand. The test is whether the debt remains serviceable if growth fails to meet expectations.

Emerging market corporates: stronger fundamentals, wider spreads

Another bright spot continues to be emerging market (EM) corporates. EM corporate balance sheets are materially stronger than those of their developed-market peers, and investors are being paid more to own them. EM’s spread premium owes more to geography than to underlying credit quality. The market embeds a premium for sovereign, political, currency and broader macro risks, even for companies with strong fundamentals. Outright valuations are not cheap, however, and dispersion remains elevated, so investors should remain selective and highly dynamic in their allocations.

Figure 3. EM’s spread premium owes more to geography than to fundamentals

Source: BofA Global Research, as of 30 September 2026.

Problems loading this infographic? - Please click here

Higher yield, but new concentrations in investment grade

Turning to investment grade, total returns continue to be driven more by interest-rate moves than by changes in credit spreads. Yields have risen across the curve, while the option-adjusted spread on the ICE Global Large Cap Corporate Index is close to where it began the year and well below its historical average. With spreads this tight, however, and rates likely to remain higher for longer, there is limited scope for further compression.

Higher US Treasury yields offer some encouragement, if history is our guide. Based on our analysis, when the 10-year US Treasury yield exceeded 5.25%, subsequent 12-month returns on USD-hedged global investment-grade credit were positive in every observation and averaged 8.5%.⁷

Figure 4: Historical global investment-grade returns 12-months forward from starting Treasury yield

Source: 12-month return of ICE Global Large Cap Corporate Index USD Hedged plotted against starting month-end US Treasury 10-year yield. Month-end data from 1 January 1999-15 September 2026. Past performance is not indicative of future results.

Problems loading this infographic? - Please click here

What that historical pattern may not fully capture is the market’s changing composition, particularly the scale of AI-related financing. Hyperscalers are estimated to spend up to US$800 billion on AI infrastructure this year,⁸ funded through a range of sources, including corporate bonds and asset-backed structures, with substantial bond supply at the long end of the curve. In contrast to the broader index, spreads on the JPM USD IG Hyperscaler Index reached approximately 140 basis points at the end of August 2026, around twice their level at the start of 2025.

Figure 5: Hyperscaler investment-grade spreads double their early-2025 level

Source: JPM USD IG Hyperscaler Index, as at 14 September 2026.

Problems loading this infographic? - Please click here

For now, that widening may largely reflect the volume of debt investors are being asked to absorb. Over time, the more consequential question is whether the investment produces the cash flows expected of it. If it does not, hyperscalers could become a significant source of pressure on investment-grade returns, especially if their share of the investment-grade index reaches the projected 10% within four years.⁹

Notably, that exposure is not confined to US dollar portfolios. Alphabet has borrowed roughly US$60 billion across six currencies,¹⁰ showing how the same issuer can appear in portfolios diversified by currency or geography. Further, pools that once behaved independently are increasingly sharing the same underlying credit risks, blunting the diversification benefits that flexible, global investors have historically come to expect.

So where are we?

In our view, the credit cycle has not reached the point where aggregate losses or defaults dominate the picture. Income is still doing considerable work, stronger issuers retain flexibility, and higher yields have improved the income available in investment grade. However, we believe the margin for error has narrowed.

In private credit, unrealised losses, non-accruals and recoveries merit closer attention even while realised losses remain contained. In high yield, tight compensation sits alongside gradually rising leverage, while floating-rate loan borrowers are more exposed to sustained high financing costs. In investment grade, investors must weigh that higher income against duration risk and the growing concentration of AI-related supply.

We believe the opportunity lies in distinguishing borrowers and structures that may carry their debt through a more demanding phase of the cycle from those priced as though nothing will change.

1. Source: https://www.ft.com/content/e485a228-1efe-426b-addc-26069ba48bf3?syn-25a6b1a6=1
2. Source: Cliffwater 2026 Q2 Report on U.S. Direct Lending.
3. Source: JP Morgan as of 30 September 2026.
4. Source: Bloomberg, ICE BofA, as of 30 September 2026.
5. Source: BofA, Bloomberg, as of 30 September 2026.
6. Source: With Intelligence, S&P Global as of January 2026
7. See Figure 3 for source information. Past performance is not indicative of future results.
8. Source: https://www.goldmansachs.com/insights/articles/global-investment-is-forecast-to-exceed-1-trillion-in-2026
9. Source: Can credit markets absorb the AI buildout? | J.P. Morgan Asset Management
10. Source: $300 Billion AI Debt Binge Spreads From Wall Street to Tokyo - SWI swissinfo.ch