Structured credit works as long as losses remain dispersed and manageable over time. When exposure becomes concentrated in one sector, that protection weakens — especially if the risk is structural rather than company-specific. That’s what’s emerging in software loans: a move from isolated weakness to a more correlated stress across the structure.

On March 17, CLO managers were reported to be cutting exposure to software loans amid concerns over downgrades, weaker cash flows, and AI-related disruption. Software represents a large share of CLO collateral, making the shift more significant. Spreads widened, deal activity slowed, and managers became more selective.

This followed earlier signs of stress, including redemption limits at private credit funds and loan markdowns tied to software exposure, as lenders reassessed risk and tightened conditions.

A CLO is a structure that redistributes risk — losses hit junior tranches first, protecting senior investors as long as defaults stay contained. That works with dispersed risk, but not with concentration. If many loans rely on the same underlying assumption, diversification weakens.

In this case, software was treated as stable and less cyclical. The current stress suggests that assumption may be more correlated than expected, especially if AI reshapes the sector broadly.

CLOs don’t remove risk — they distribute it. Equity takes first loss, mezzanine follows, and senior debt is protected, while investors and lenders each absorb pressure in different ways.

In stable conditions, this looks orderly. Under stress, the structure becomes visible: losses move through tranches, liquidity tightens, and leverage is reduced. Risk isn’t gone — it’s transferred across balance sheets.

Structured credit depends on liquidity — until it doesn’t. When many investors try to exit at once, prices adjust quickly and discounts widen. That creates the tension: preserving value means limiting forced sales, but that often requires restricting liquidity or tightening financing. The system cannot fully provide both stability and flexibility at the same time.

What persists: the need for non-bank credit. Companies still require financing, and structured credit still offers a way to distribute lending risk across different types of investors. That basic function is not disappearing.

What can break: the assumption that software is automatically defensive collateral. That was a useful idea while the sector looked stable, recurring, and less exposed to abrupt disruption. It becomes weaker if the same technological shift starts challenging revenue durability across many borrowers at once. The fragile part is not the existence of software lending. It is the belief that a large concentration in software still behaves like broad diversification.

The recent selling in software loans is structurally relevant because it tests one of structured credit’s quiet requirements: weakness must stay dispersed enough for the design to work. A CLO can absorb defaults. It is less comfortable absorbing a broad challenge to the same underlying earnings story across a large part of its collateral.