Much of the debate about artificial intelligence and software as a service focuses on whether AI will make existing applications obsolete. From a credit perspective, we believe that question is too broad and AI does not need to make a software company obsolete to create a credit problem. In our view, it only needs to weaken the economics that historically made software attractive: high retention, pricing power, strong gross margins and predictable recurring revenue. Worldwide software spending is forecast to grow more than 15% next year, and a market expanding that quickly can still contain businesses whose credit quality is deteriorating.
Look beneath revenue growth
Annual Recurring Revenue (ARR) can continue to grow through new logo bookings even as net retention declines, expansion slows or existing customers contract. The mix is important: growth that increasingly depends on replacing ARR with new customers may be less attractive than growth driven by a healthy installed base.
In our view, examining renewal patterns across comparable groups of customers can help reveal whether the existing revenue base is becoming more or less stable. We believe the reasons behind those patterns may matter just as much. A delayed renewal could reflect a customer’s procurement process rather than dissatisfaction with the product. Lower spending may reflect fewer seats or lower usage at the customer rather than weakening demand for the product itself. In our opinion, those distinctions can help show whether pressure is customer-specific or a sign of a broader change in demand.
One widely covered example from the past two years illustrates why that context is important. In the summer of 2024, Klarna chief executive Sebastian Siemiatkowski announced that the company was shutting down Salesforce and Workday as part of a sweeping AI overhaul, prompting questions about whether AI could accelerate displacement of established enterprise applications. The reality was more nuanced. Within months he was clarifying that Klarna had consolidated its systems around an internally built data stack rather than replacing its software with a large language model. We believe the competitive risk for established vendors may come not only from other software providers, but from customers becoming better able to internalize functionality they previously purchased.
This is also where the discussion about software’s defensibility becomes practical. Mission-critical and defensible are not necessarily the same thing. If AI makes implementation, data migration or customization easier, in our view, switching costs may decline even when the underlying workflow remains essential. The first signs may appear in lower net retention, seat contraction, reduced usage or greater pricing concessions well before customers churn outright.
For lenders, those shifts can pressure margins and enterprise value before the impact is obvious in reported revenue.
Connect operating changes to cash flow
AI may improve a software company’s economics by reducing the cost of serving customers or adding capabilities they are willing to pay for. It can also increase the cost of maintaining competitive parity. Customers may come to expect AI functionality as part of the core product, while vendors absorb higher development, compute and infrastructure costs.
Consider a company that adds an AI feature to a fixed-price subscription but pays an outside technology provider each time that feature is used. Greater adoption could make the product more useful to customers while increasing the company’s costs. Strong usage would be encouraging, but it would not, on its own and in our view, establish that the business is generating more cash.
If the economics do not support that spending, higher development and infrastructure costs can pressure gross margins and cash generation, particularly for borrowers with limited financial flexibility.
Repayment paths matter
In our view, for many growth-stage software borrowers, repayment may not be expected to come solely from deleveraging through free cash flow. The debt may often ultimately be taken out through a refinancing or a sale of the company. That makes the availability of capital and the market’s view of enterprise value part of the credit story, even when the borrower remains current on its obligations.
In our view, traditional exits have become more complicated. When lenders and acquirers are both trying to determine how AI may change a software company’s moat, pricing power and terminal value, we believe the two most common paths to repayment could narrow at the same time.
The timing is notable. In the broader software credit market, PitchBook/LCD data point to roughly $40 billion of software loans maturing in 2028, with more than half of that cohort rated B-minus or lower. A meaningful share of those loans were originated in 2021, when rates were low, software valuations were elevated and lenders had greater confidence in recurring revenue. The risk is that these borrowers may arrive at maturity with less lender appetite, lower valuation support and fewer sale options than the original capital structure assumed.
This is where AI can become a balance-sheet issue before it becomes an operating crisis. A company can remain covenant-compliant and continue to grow while the value supporting the debt is declining. If uncertainty around AI compresses valuation multiples or shrinks the universe of potential buyers, we believe the cushion beneath the loan can erode before liquidity does.
From our perspective, AI does not justify dismissing software as a category. Nor does recurring revenue, by itself, establish that a borrower is insulated from change. In our view, the more useful question is whether AI strengthens or erodes the characteristics that made a particular software business financeable in the first place. Some companies will use AI to deepen their moat and strengthen cash generation. Others may discover that recurring revenue can remain in place even as valuations move lower, making refinancing and sale outcomes more fragile.
Avisha Khubani is Chief Credit Officer at Runway Growth Capital.
