Analysis: UBS Warns AI Disruption Is Spreading Into Credit Markets, Forecasting Up to $120 Billion in Defaults

To understand why UBS is sounding the alarm on credit markets, one need only look at the devastation already visited upon publicly traded software companies. What began as isolated sell-offs has metastasized into a sector-wide reckoning, driven by a fundamental reassessment of which businesses can survive the AI transformation — and which cannot.

The numbers are staggering. As of February 17, 2026, Salesforce (CRM) trades at $184.47, down 44% from its 52-week high of $329.74. ServiceNow (NOW) has been cut nearly in half, falling to $105.43 from a high of $211.48 — a 50% decline. Workday (WDAY) sits at $143.17, down 49% from its $281 peak, touching new 52-week lows in today's session. Even cybersecurity leader Palo Alto Networks (PANW) has fallen 27% from its highs to $163.14, while Microsoft — the world's largest software company — has shed 28% of its value, dropping from $555.45 to $398.63.

Polar Capital, which manages a $12 billion global technology fund that had outperformed 99% of its peers, has reportedly begun dumping software holdings as the iShares Expanded Tech-Software ETF has fallen 22%. Nick Evans, the fund's manager, is making what he calls a "decisive call on the future of software" — and that future, increasingly, looks bleak for companies that can't rapidly integrate AI into their core offerings. The market is shifting from viewing AI as a rising tide for all technology companies to a winner-take-all dynamic where foundational model makers like Anthropic and OpenAI threaten to displace incumbents entirely.

While public equity markets have been quick — if brutal — in repricing software risk, the leveraged loan and private credit markets have been far slower to react. That, according to UBS's Mish, is precisely what makes them so dangerous.

The combined leveraged loan market ($1.5 trillion) and private credit market ($2 trillion) are home to thousands of below-investment-grade companies, many of them private equity-backed software and data services firms carrying substantial debt loads. Mish estimates that default rates in leveraged loans could rise by up to 2.5 percentage points, while private credit defaults could jump by up to 4 percentage points by late 2026. Applied to the total market size, that translates to $37.5 billion to $60 billion in leveraged loan defaults and $40 billion to $80 billion in private credit defaults — a combined $75 billion to $120 billion in fresh defaults.

"The market has been slow to react because they didn't really think it was going to happen this fast," Mish told CNBC. "People are having to recalibrate the whole way that they look at evaluating credit for this disruption risk, because it's not a '27 or '28 issue." The speed of AI model improvements has compressed what many investors assumed would be a multi-year transition into something that could unfold over quarters, not years.

Perhaps more troubling is Mish's tail-risk scenario, in which defaults jump to twice his base estimates. In that world, the knock-on effects cascade through the financial system: funding dries up for many companies, credit spreads blow out across all leveraged debt, and what begins as a technology disruption story becomes a systemic financial stress event. "You will have a broad repricing of leveraged credit, and you will have a shock to the system coming from credit," Mish warned.

UBS's framework divides the corporate landscape into three tiers — and investors' ability to distinguish between them may determine who navigates the coming credit storm and who gets caught in it.

The first tier consists of the foundational AI model creators: Anthropic, OpenAI, and a handful of others building the large language models that are reshaping entire industries. These companies are currently private but could soon be among the largest publicly traded firms in the world. They represent the source of the disruption, not its victims.

The second tier comprises investment-grade software giants with robust balance sheets and the resources to integrate AI into their existing products. Companies like Salesforce, Adobe, and Microsoft fall into this category. While their stocks have been punished — Salesforce now trades at 24.6x earnings with $11.3 billion in cash on its balance sheet and a modest 0.19x debt-to-equity ratio — they have the financial firepower to adapt. Salesforce continues to generate over $10 billion in quarterly revenue, with its most recent quarter showing $10.26 billion in sales and a 78% gross margin. The company's Agentforce AI initiative and its upcoming Q4 earnings report on February 25 will be a critical test of whether enterprise AI spending is translating into real revenue.

The third and most vulnerable tier is the cohort of private equity-owned software and data services companies — firms with high debt loads, limited public visibility, and business models that may be rendered obsolete by AI. "The winners of this entire transformation — if it really becomes, as we're increasingly believing, a rapid and very disruptive or severe change — the winners are least likely to come from that third bucket," Mish said. These are the companies that populate the leveraged loan and private credit markets, and they are the epicenter of UBS's default forecast.

Adding complexity to the outlook is the disconnect between monetary policy easing and rising credit stress. The Federal Reserve has steadily lowered its benchmark rate over the past year, from 4.33% in February 2025 to 3.64% in January 2026, representing 175 basis points of cumulative cuts. The 30-year mortgage rate has edged down modestly to 6.09%, and the 10-year Treasury yield has drifted lower to 4.09% as of February 12.

But lower rates cannot solve the fundamental problem UBS is identifying. When a company's core business model is being disrupted by AI — when the software it sells can be replicated or surpassed by a large language model at a fraction of the cost — cheaper borrowing costs are at best a palliative, not a cure. The 2-year Treasury yield at 3.47% and a positively sloped yield curve (the 10Y-2Y spread stands at 0.62%) suggest the bond market sees continued easing ahead, but that easing is calibrated for a cyclical slowdown, not a structural technology disruption.

The labor market adds another layer of uncertainty. Unemployment ticked up to 4.3% in January 2026, the second consecutive monthly increase after hitting 4.5% in November. If AI-driven defaults begin to cascade through leveraged companies — many of which are significant employers — the intersection of technology disruption and labor market weakness could create a feedback loop that monetary policy is ill-equipped to address.

Several catalysts in the coming weeks will help determine whether UBS's warning proves prescient or premature. Salesforce reports Q4 earnings on February 25, with analysts estimating approximately $3.41 per share in EPS on roughly $12 billion in revenue for the April quarter. The results — and more importantly, management's commentary on AI-driven enterprise spending — will serve as a bellwether for the broader software sector's ability to navigate the transition.

Workday reports on February 24 as well, providing another critical data point. At 60x earnings and down 49% from its highs, the market is already pricing in significant pain — but guidance will reveal whether the deterioration is stabilizing or accelerating.

Beyond individual earnings, investors should monitor high-yield credit spreads, leveraged loan prices, and private credit fund redemption requests for early signs of the stress Mish is forecasting. The gap between public equity repricing (which has been swift and severe) and private credit repricing (which has been sluggish) represents either a rational assessment that private markets are more insulated — or a dangerous lag that could close violently.

The key variable, as Mish acknowledges, is the pace of AI adoption by large enterprises. If companies accelerate their shift from traditional software licenses to AI-native solutions — as the latest capabilities from Anthropic and OpenAI make increasingly plausible — the timeline for credit distress compresses further. "We're not yet calling for that tail-risk scenario, but we are moving in that direction," Mish cautioned. For investors in leveraged credit, the time to stress-test portfolios for AI disruption risk is not next year. It is now.