Software loans: better prices, unchanged economics
Software has been one of the weakest segments of the US leveraged loan market in 2026. The trigger was the concern that AI could automate the very workflows many software products charge for. Because software accounts for more than 10% of the Morningstar LSTA US Leveraged Loan Index, the sell-off has weighed on the entire asset class.
PIMCO's latest Credit Market Lens, published on 19 October, describes a recovery that has stalled. Spreads tightened over the summer, retracing about one third of the year's widening, but have been widening again since early September. The average bid price of software loans remains about 8 points below that of non-software loans, compared with around 10 points in June.
«The AI disruption narrative that catalyzed this year's sell-off in software loans remains unchanged», writes Lotfi Karoui, Multi-Asset Strategist at PIMCO. A small group of issuers appears well positioned to benefit from AI. A much larger group faces uncertain competitive positions, and a growing share contends with business model and execution risks.
The next test is refinancing. About USD 62 billion of software loans in the index mature in 2028 and 2029. Recent amend-and-extend deals show how differently the market prices borrowers: some extended maturities for an additional 25 basis points of spread, others paid 150 basis points more and accepted an issue price below par. PIMCO's conclusion: «For weaker credits, spread compression has improved the optics more than the economics.»
PIMCO also warns against reading the rebound in software equities as relief for software credit. The large listed software companies are often investment grade issuers with broad access to capital markets. Leveraged loan borrowers do not enjoy the same financial flexibility.
Recurring revenue is not the same as durable revenue
Software lending rested on a simple assumption: subscription income is predictable, so it can support high leverage. AI calls that assumption into question. If customers need fewer licences, or if an AI tool replaces a specialist application, retention and pricing power erode, and with them debt capacity.
The relevant credit question is therefore not whether a borrower can pay today's coupon. It is whether its subscription cash flows will remain defensible long enough to repay or refinance the debt. A maturity extension buys time. It does not repair the economics.
How a sector problem travels through global fixed income
The stress does not stay within US loan indices. Software is a significant share of direct lending portfolios, and its weakness has added to redemption pressure in semi-liquid private credit vehicles, as investrends.ch reported in October. Three channels matter for international investors:
- Leveraged loans and CLOs: a diversified loan allocation can still carry a concentrated business model risk if many borrowers depend on the same assumptions about subscription growth.
- Private credit funds: loans without a market price are valued by model. Stable reported values should be read alongside cash interest versus payment-in-kind interest, amendments and borrower performance.
- Pension and insurance portfolios: international funds and securitisations carry the risk to investors well beyond the borrowers' home markets.
The other side: Switzerland as a borrowing venue
While leveraged software borrowers struggle to refinance, the AI infrastructure builders are borrowing at scale, and the Swiss franc market has become part of their funding mix.
In February 2026, Alphabet placed an inaugural CHF 3.055 billion bond in five tranches, a record debut in the Swiss franc market. Amazon followed in May with CHF 2.82 billion in six tranches, a record for the currency, with maturities from three to 25 years. According to research by Zurich Insurance cited by the Financial Times in September, AI-related companies accounted for 26.4% of Swiss franc corporate bond issuance this year.
The attraction goes beyond a low coupon. The Swiss market offers a deep base of long-term buyers, namely pension funds, insurers and bond funds, as well as reliable execution and a way for issuers to diversify away from US dollar investors. Whether a CHF bond is actually cheaper for a dollar-based issuer depends on the currency hedge, not on the headline coupon.
Two AI credit risks in one portfolio
For Swiss institutional investors, this creates two distinct exposures:
- Lending to companies whose products may lose value because of AI, via leveraged loans, CLOs and private credit funds.
- Lending to companies building AI infrastructure, whose risk is whether very large investments earn adequate returns, now increasingly via CHF bonds in the domestic portfolio.
The two risks differ, but conventional asset class labels hide both. KKR estimates AI-related debt at around USD 600 billion, or 6.3% of the US investment grade market, before leases and guarantees. A CHF denomination does not turn a US hyperscaler into a Swiss economic exposure. New CHF bonds broaden the domestic issuer universe; they can also add a further claim on the same technology ecosystem already dominant in global equity portfolios, a point investrends.ch examined in July.
Questions for investment committees
- How much software exposure remains in loan, CLO and private credit allocations after looking through sector labels?
- What share of reported interest income in private credit holdings is paid in cash rather than in kind?
- Have the new hyperscaler issues in the Swiss franc market changed issuer limits in the CHF bond portfolio?
- Is AI-related risk measured across equities, bonds, private credit and infrastructure as a whole?
The bottom line
The software loan market is increasingly defined by winners, losers and wide dispersion. Selective lending, rather than a sector call, is the logical response. At the same time, the Swiss franc market is becoming a funding source for the AI build-out. Both developments call for the same discipline: measure AI exposure at total portfolio level, not asset class by asset class.
Disclosure: PIMCO is a partner of the Swiss Institutional Symposium 2026.
Sources: PIMCO, Lotfi Karoui, «The Credit Market Lens: Improved Optics But Unchanged Economics in Software Loans» (19 October 2026, data as of 14 October 2026); Homburger (Alphabet CHF issue, February 2026); Bloomberg via swissinfo.ch (Amazon CHF issue, 12 May 2026); Financial Times, September 2026, citing Zurich Insurance research; investrends.ch (23 July, 27 August and 1 October 2026), incl. KKR estimate. This article does not constitute investment advice.