Analyst sends stark $120 billion AI warning | Official Website of Louis Velazquez, entrepreneur, finance guy and tech innovator

Everyone’s been asking the same question about artificial intelligence: who’s going to win? Chipmakers have soared. Software companies have rallied. Investors have bought into the idea that we’re at the start of a major productivity boom.

But Matthew Mish atUBS is asking something else entirely. He wants to know who’s going to lose, and more importantly, how much damage they’ll do on the way down.

Mish runs credit strategy at UBS, which means he spends his days thinking about corporate debt. And what he’s seeing has him worried. It’s not about whether AI works. He thinks it probably will. The problem is speed.

AI isn’t rolling out gradually anymore. It’s moving from labs into actual businesses faster than many expected. That’s great if you’re ready for it. If you’re not, and you’ve got a pile of debt on your balance sheet, things could get ugly pretty quickly.

Photo by Erik McGregor on Getty Images

Companies don’t have years to figure this out anymore

Here’s the thing about credit markets. When banks and investors lend money, they assume the borrower will have time to adjust if competition heats up or technology changes. Maybe three years. Maybe five.

But if AI compresses that timeline to 18 months or less, many of those assumptions break. Revenue can drop faster than anyone planned for. Cash flow projections that looked solid six months ago suddenly don’t hold up.

More AI Stocks:

Mish told CNBC that markets have been “slow to react because they didn’t really think it was going to happen this fast.” He’s talking about what he calls a “rapid and aggressive disruption scenario,” which is Wall Street speak for things happening too fast for people to get out of the way.

The numbers are big

UBS has been running the math. They’re looking at about $3.5 trillion in leveraged loans and private credit that could be sitting in the blast zone. These aren’t the safest borrowers. They’re companies with weaker balance sheets, higher debt loads, and less room for error.

In UBS’s baseline scenario, defaults could hit $75 billion to $120 billion by the end of this year. Default rates could climb to around 2.5% for leveraged loans and as high as 4% for private credit.

Those are manageable if everything else stays calm. But if disruption picks up speed, Mish thinks defaults could go twice as high. At that point, you’re dealing with a repricing of risk across the entire credit market.

Higher risk means higher borrowing costs. Higher borrowing costs mean less lending. Less lending means companies that need to refinance debt are going to have a harder time, even if their own businesses are fine.

Software companies are right in the crosshairs

Mish has zeroed in on lower-quality software firms as particularly vulnerable. A lot of these companies are built on subscription revenue and borrowed money. They’ve been doing fine in a world where their software solves specific problems and customers don’t have cheaper alternatives.

That world is changing fast. Coding tools, data analysis, customer service bots. A lot of what used to require specialized software is getting bundled into bigger platforms or replaced by AI that does the same job for less money.

If you’re a weaker software company watching your pricing power erode and your customers start to churn, you’ve got a problem. Your revenue assumptions no longer work, but your debt payments haven’t changed.

Mish’s argument is straightforward. The companies winning big from AI aren’t the ones borrowing heavily in the leveraged loan market. The ones that are borrowing heavily are getting squeezed between technological change they can’t control and debt obligations they can’t escape.

It’s not about everything falling apart overnight. It’s about enough marginal companies hitting trouble at the same time that the credit system starts to buckle.

What happens next depends on how fast this moves

For stock investors, this adds a wrinkle to the AI story that hasn’t gotten much attention. Most of the debate has been about whether Nvidia is too expensive or whether Microsoft can justify its valuation. Mish says the real stress might first show up in bond markets.

If defaults start climbing in leveraged loans and private credit, it’s going to make investors nervous across the board. That means less appetite for risk. That means companies that rely on borrowed money are going to feel pressure, even if AI demand itself stays strong.

It also raises questions about private credit funds, which have grown rapidly over the past few years as banks backed away from riskier lending. These funds have been marketed as steady income generators. Higher defaults would test that pitch quickly.

In the wo rst case, tighter credit spills over into the real economy. Smaller companies slow down hiring. Investment gets pulled back. Eventually, that feeds into corporate earnings and economic growth, and suddenly you’re dealing with something bigger than just a few troubled borrowers.

Mish hasn’t called this a sure thing. He’s been careful to frame the worst outcomes as tail risks, not his base case. But he’s also said his team is “moving in that direction” as AI reshapes business models faster than anyone thought possible just a year ago.

AI is supposed to make things more efficient and more profitable. Mish’s warning is a reminder that it also makes things more fragile. Companies that can’t keep up no longer get a grace period. And when enough of them stumble at once, it’s not just their problem.

The next chapter of the AI story might not be about which stocks go higher. It might be about whether the credit system can handle the speed of what’s coming.