Morgan Stanley: AI Risk Is Widening Software Loan Price Gaps as Refinancing Pressure Emerges Before Earnings Impact

Which software loans face higher AI risk?

Of the 247 billion dollars in software loans covered by Morgan Stanley, 16% by outstanding balance are classified in the red group at high risk of AI disruption, 49% in the medium-risk yellow group, and 35% in the low-risk green group. Of rated loans in the red group, 72% are rated B- or lower.
The bank assesses both the moat around existing businesses and their ability to adapt to AI. The former considers whether products are embedded in critical workflows and difficult to replace; the latter considers whether technology, pricing and funding can support the transition. Red loans account for 32% of front-office application loans, above the overall 16%; back-office and vertical applications are more resilient because of factors such as critical business systems, compliance and proprietary data.
The assessment relies mainly on public information and does not explicitly incorporate private equity shareholders’ ability to manage portfolio companies. Changes in corporate strategy and technology could also alter the risk scores.

As of 29 October 2026, the weighted average bid prices for green, yellow and red loans were 97, 88 and 73, respectively. The green group recovered about half its decline from the March low, while the other two groups rebounded less.
Morgan Stanley’s regression shows that, after controlling for ratings and maturities, each 10-point increase in the AI risk score is associated with a loan price about 5 points lower on average. However, red loan prices vary widely: bid prices for the middle 50% range from approximately 50 to 94. Leverage, liquidity and individual restructuring events also affect valuations.

Of approximately 39 billion dollars in red loans, around 30 billion dollars mature in 2027–2029. Morgan Stanley estimates that if low-B and CCC-rated red borrowers refinanced debt maturing before 2029 under market conditions at the time, coupon spreads could rise by 550–1000 basis points from existing levels.
Funding channels remain open but are more selective: of the 11 completed refinancing or extension transactions in software and related sectors, 10 were in the green or yellow groups. Recent extensions typically secured terms of 1–3 years, but spreads rose by around 50–225 basis points. Recent transactions have also led the bank to believe that default outcomes may be better than it expected at the start of the year.
Operating data remain resilient for now: software borrowers’ EBITDA grew 5% year on year in the second quarter of 2026, below the 8% for the broad leveraged loan index; this software sample contains only 10 issuers. The bank believes constrained financing could accelerate credit deterioration among weaker borrowers before AI disruption is fully reflected in operating performance.

Of the 133 billion dollars in software loans held by CLOs covered by the bank, the red group accounts for 12%, below the 16% for the overall software loan market. Since February, managers have collectively made net purchases of 2.8 billion dollars in green loans and net sales of 1.1 billion dollars in red loans, while shifting toward assets with later maturities.
The adjustment is not cost-free: selling loans at a discount and buying higher-priced, lower-risk assets can crystallize losses in the face value of the collateral pool. Morgan Stanley also finds that CLOs with a higher share of red loans tend to have lower BB-rated market value coverage and wider secondary spreads; other collateral characteristics and transaction structures also influence this relationship.
The bank therefore remains cautious on the lower tiers of the CLO capital structure, emphasizing selection of assets and managers based on borrower quality, maturity profiles, loan valuations and transaction structures.