Lenders now ask the AI question too

For the owner of a software company that will refinance or be sold in the next two years, the lender's questions now sound much like the buyer's. PitchBook's LCD reported on October 2 that software companies accounted for 12% of private equity-backed direct lending deals by count so far in 2026, down from 16% last year, and for an estimated 15% of volume, down from 22%.^[1] Lenders whose portfolios are full of software loans have grown pickier, it said, and most software deals this year have been refinancings and extensions by the lenders already in place.^[1] Our conclusion: prepare the evidence on AI exposure for a refinancing as carefully as for a sale, and start before the maturity date sets the timetable.

The maturities are arriving

LCD counted a record $117 billion of loans at business development companies due within two and a half years as of the second quarter, about 22% of their holdings, and said 21% of that amount comes from software businesses.^[2] S&P Global Ratings expects software maturities at the business development companies it rates to stay manageable through 2027 and to rise to about $3 billion in each of 2028 and 2029.^[3] New money is scarcer. KBRA DLD has reported five new software loans of $1 billion or more this year, against 18 in 2025.^[4]

The picture is not uniformly weak. S&P found that more than 80% of the software companies it follows through credit estimates increased revenue from 2025 to 2026.^[3] The Bank for International Settlements found in July that software loans at business development companies face no nearer-term wall of maturities than other sectors.^[5] The pressure falls on access and terms more than on current results.

What lenders now ask

KBRA reported in March that it had seen lenders require additional covenants, raise spreads by as much as 100 basis points or more and, in some cases, decline to extend maturities for underperforming companies.^[6] It said lenders were seeing early signs of change driven by AI in customer behavior, longer sales cycles and budgets moving toward AI initiatives.^[6] Where software loans are amended and extended, S&P expects lenders to ask for some combination of higher coupons, partial paydown, tighter covenants and documentation, or more equity.^[3]

The questions are the ones buyers ask. S&P judged each company's exposure by the function its product performs, how deeply it is embedded in customer operations and how easily it could be automated or replaced; it placed 32% of the software companies it reviewed at some risk or more, and 25% as neutral or positive.^[3] Moody's research in September said investors increasingly want proof of durable cash generation rather than simply recurring revenue, and that outcomes for borrowers in the same sector depend on earnings quality, customer concentration, leverage and valuation support.^[7]

Bring the evidence to the lender

Give the lender the analysis a buyer would ask for. Show revenue by product and by the job each product does for the customer, how deeply it sits in customers' workflows, and which parts could be automated or replaced. Report gross retention beside net retention by cohort, revenue by pricing model and the concentration of the largest customers. Where AI is already in the product, show what customers pay for it and what it does to gross margin.

Then show the cash. Reconcile EBITDA to free cash flow over recent quarters, and test interest cover at a wider spread and under tighter covenants than the current loan carries. Decide in advance what the company could offer if asked, whether a partial paydown, more equity or a shorter term, and what it would not accept.

Start well before the final year of the loan, so that an amendment remains a choice rather than the only route. A lender that sees the evidence early can judge the company on its own record rather than on the sector's, which is the same reason to prepare it for a buyer.