JP Morgan's AI Fears Trigger Private Credit Revaluation. Liquidity Risks Loom.

JP Morgan has begun writing down the value of software company loans held by private-credit firms, cutting the amount it will lend against these assets on concerns that artificial intelligence could undermine the businesses underpinning hundreds of billions in debt. The move represents the first time a major Wall Street bank has adjusted its risk assessment based explicitly on AI displacement fears. For an industry built on careful actuarial calculations rather than technology trend forecasts, this is a striking development.

JP Morgan isn't hedging or issuing cautious notes to clients—it's marking down actual asset values on its books. Chief executive Jamie Dimon has repeatedly drawn parallels between current market conditions and the months preceding the 2008 crisis. At the bank's investor day in February, he noted he was seeing "people doing dumb things"—language consistent with his historically cautious public stance, though this time the perceived threat comes from technology rather than subprime mortgages.

What makes Dimon's warnings particularly noteworthy is that JP Morgan was among the few institutions that spotted mortgage risks early in 2007, giving his concerns a credibility that tends to reverberate through trading floors. The immediate question is whether those fears are warranted. There's no concrete evidence yet that AI will replace the specific software companies whose loans are now being devalued, only that lenders believe it might.

Enjoying this article?

Get stories like this in your inbox every week.

Join the Community

But in credit markets, perception can become reality fast. If banks won't lend against these assets, their liquidity evaporates regardless of whether the underlying businesses remain profitable.

A £1.6 trillion market faces its first real stress test

Private credit has swelled to roughly $2 trillion—£1.6 trillion—over the past decade as pension funds, insurance companies, and wealthy individuals chased higher yields outside traditional banking regulation. These loans offer returns several percentage points above public debt, making them attractive to funds that need to meet ambitious return targets. The catch is that private credit instruments don't trade on public markets.

Unlike corporate bonds that can be sold within seconds, private loans have no ready buyers. That illiquidity isn't a problem when everyone wants in, but it becomes one quickly when investors head for the exits. That stress is playing out in real time.

Blackstone's $82 billion BCRED fund saw investors attempt to pull 7.9 per cent of shares last week—roughly $3.8 billion—a record redemption request. The fund typically limits quarterly withdrawals to five per cent, forcing Blackstone and its own employees to inject capital to meet the extra 0.9 per cent gap. Days later, BlackRock received requests to redeem 9.3 per cent from its $26 billion HPS Corporate Lending Fund and simply enforced its five per cent cap, leaving investors with roughly half the cash they requested.

The mechanics here matter. When redemption requests exceed limits, investors can't access their capital even if they urgently need it. That's a fundamentally different proposition from a publicly traded bond fund, where you might take a loss but can exit same-day.

The structure works brilliantly during calm markets and creates panic during stress.

From Silicon Valley buzzword to balance sheet risk

What's particularly interesting is how quickly AI has moved from technology sector talking point to concrete financial risk assessment. Just last month, anxiety intensified after Anthropic launched a new AI tool, coinciding with sharp falls across European software stocks—though attributing market movements to single causes is always speculative, and those losses may not prove sustained. But banks don't write down assets based on market volatility alone.

JP Morgan's decision suggests its credit analysts believe certain software businesses face genuine structural threats from AI tools that can perform similar functions at lower cost. The bank isn't making a blanket call on technology—it's specifically targeting software firms whose business models might be vulnerable to displacement. Legacy enterprise software companies, which often charge substantial licence fees for relatively straightforward functions, could indeed face pressure from AI agents capable of handling similar tasks.

Whether that translates to actual defaults depends on how quickly businesses adopt new tools, how effectively incumbent software firms adapt, and whether AI capabilities live up to current expectations. The challenge for private-credit investors is that their portfolios are disproportionately weighted toward exactly these kinds of businesses.

Software companies were attractive loan candidates precisely because they had recurring revenue, high margins, and sticky customer relationships. If those advantages erode, the collateral underpinning billions in loans weakens. Pension funds and insurance companies holding private credit assets through fund investments may not realise the concentration risk they've taken on.

The sector's opacity—loans aren't publicly reported or priced—means many institutional investors don't know exactly which companies their money is financing. They'll discover the exposure when funds report losses or, worse, when redemption requests can't be met. JP Morgan's markdown may prompt other banks to reassess their own lending against private credit assets, potentially tightening liquidity across the sector.

If that happens, private-credit funds will find it harder to finance new deals and may need to hold more capital in reserve, reducing returns. That would test investor appetite at precisely the moment redemption pressures are rising, creating the conditions for a self-reinforcing pullback. UBS has warned of a rapid AI shock triggering sharp rises in defaults across the US credit market, while JPMorgan executives have expressed surprise at market shock over the private-credit meltdown.

More articles by Ross Williams