A Howard Marks memo, the kind all of Wall Street reads. This one's subject: the first credit accident caused by AI. The backstory: for a decade, private equity treated lending to software companies as the safest trade around. Subscription revenue arrives like clockwork, so how could they fail? Software grew to 20-30% of the private lending market on that logic. Then AI made writing code dramatically cheaper, and suddenly a question appeared: do these companies still have a moat? In February, investors all asked it at once. Withdrawal requests piled up, funds gated redemptions, and the gating made everyone more nervous. Marks's diagnosis is calmer than the panic: most of the borrowers are actually fine, and this is a sentiment problem, not a solvency one. But he notes the whole sequence matches, beat for beat, the bubble pattern he's been writing about for decades: something new, a grain of truth, early winners, envy, then latecomers with low standards.
Why it matters · This is the first big memo treating AI not as a stock story but as a credit event. If AI can rewrite what a company is worth, it rewrites what its debt is worth.