In the Road Runner cartoons, the coyote runs on past the cliff’s edge and only begins to fall once it looks down. Richard Rumelt borrows that moment in Good Strategy Bad Strategy to explain credit bubbles, and it fits a public market just as well. The edge arrives the same way: a company posts growth projections until the numbers on the page no longer cover the bills behind them.
Intel’s stock fell to its lowest point in five decades after a forecast left AI demand unanswered (the forecast fell short), and no boom in the room could paper over that gap. What happens when the ground beneath a growth story turns out to be air?
Investors blamed Intel’s stock crash on the company alone, but Blackstone’s downgrade of Anthropic’s debt shows the same pressure reaching every leveraged AI bet. A weaker answer credited a cyclical chip market or a botched turnaround. But Anthropic’s debt re-rating says the same reckoning is working through the other side of the ledger. Blackstone now expects the lab to carry junk-grade obligations by 2027, reversing the investment-grade outlook set only months ago, the FT’s report on the downgrade, because financing a $63 billion compute bill has outpaced the revenue that could match it. Public markets are not choosing Intel’s revenue problem or Anthropic’s cost problem in isolation. They are pricing the income statement whole, and finding the growth story thinner than the arithmetic.
The assumption that has held private capital since the boom began is that AI demand will outrun every cost and justify every bet. That assumption breaks the moment levered holders step past support and discover the books no longer cover the bills. Once asset prices drift downward, holders rush to sell before their equity is wiped out, and those sales drive prices faster still. The cliff edge is not coming; it is already crossed. What falls is the gap between financed costs and collectible revenue, and the longer the tab runs, the farther the drop.
If that cliff becomes the template every AI company faces at the public-market gate, private investors will stop pricing the bet and start pricing the bill. Labs that borrowed time with a $63 billion chip tab will probably discover that financing compute is not the same as earning a return on it. Utility-grade multiples could land on any balance sheet that looks more like a power plant than a software shop. A generation of private rounds priced on narrative alone may run into a public market that divides revenue by the same per-unit costs and prices in them.
Public markets do not doubt that AI has arrived. They doubt a growth story can pay the bill.
Leave a Reply