Collateral Decay Cycle

The chain of dying collateral

Asked how AI had changed what he was willing to lend against, the credit investor Marc Rowan answered that it had not, because collateral has always been dying: "we've always had that mentality, which is change is a constant."¹ Yellow pages is his anchor example, chosen because in 2000 it was defensible on every axis a lender cares about: free at the point of use, comprehensive, and a cultural habit. He notes that in his current associate class, using the term yellow pages draws raised hands from people who want to know what it was. The chain continues from there, and each destroyer becomes the next victim in turn: television and radio franchises, once thought to be massive, were diminished by other content forms; cable television was in part their successor, and was itself replaced by satellite television, which was replaced by mobile telephony, which was replaced by fiber. Rowan is careful about the failure mode here: these businesses mostly have not disappeared, they have been diminished, which for a lender's purposes is often sufficient, since collateral rarely goes to zero, it goes to a fraction of what it was underwritten at.

Structure, not forecasting

The useful part of Rowan's argument is that he does not claim a lender should predict which asset dies next. The discipline he describes is defensive and structural: be diversified, so that no single collateral thesis can do real damage; be senior where risk is perceived, since seniority is what gets a lender paid before an impairment reaches them; look for hard collateral, meaning physical, reusable, redeployable value that survives whatever business model currently sits on top of it; and shorten the underwriting horizon, accepting that a lender can make a decision for three, five, or seven years rather than twenty or thirty.

Whether AI compresses the cycle

Rowan's own account of enterprise software credit suggests the cycle may be compressing faster than a shortened horizon can absorb. In a separate interview recorded roughly six weeks earlier, he describes the market's recognition of AI risk in enterprise software as having happened "two weeks ago," while in the later interview the same event is placed at "eight, twelve weeks ago," together dating the recognition to roughly mid-February of that year.² Yellow pages took about a decade to die. A software multiple, by his own account, can reprice within weeks once a market chooses to notice. His own framing absorbs this without quite conceding it: his complaint is not that the change was fast, but that it was late. "Lo and behold, we think that AI might impact software. Is that news? Did we just discover that two weeks ago?"² AI's impact on software was foreseeable rather than a surprise. The suggestion this leaves is that the real update to the rule about shortening horizons is closer to assuming that repricing is instantaneous once a market notices, and staying senior and diversified enough that the timing does not matter.

An adjacent frame

Palmer Luckey has described a related lifecycle pattern from the equity side: industries that decay in character before they decay in cash flow, continuing to generate revenue for years after the thing that made them defensible has already been hollowed out. Rowan's collateral chain and Luckey's framing describe the same erosion from different vantage points, one from a lender assessing what still counts as safe collateral, the other from an operator watching an industry's substance change while its top line has not yet caught up.