# The AI Boom Has a Warning System. Here's What It's Saying Now

Seven of 13 warning signals for an AI market peak are now lit, according to a Peak Watch Checklist tracking the health of the current artificial intelligence buildout. The checklist identifies specific, observable events that historically precede tech booms turning into busts.

The five largest cloud companies are on track to spend more than $750 billion on AI infrastructure this year, with over $1 trillion expected in 2027. For most of this boom, major players like Microsoft, Alphabet, Amazon, and Meta funded this spending with cash. That pattern has shifted. Spending has now outrun the cash these companies generate, forcing them to borrow.

Hyperscaler bond sales hit $159 billion by mid-year, already more than all of 2025. Alphabet sold a record $85 billion of new stock. When a boom shifts from profits to borrowed money, lenders start to matter as much as customers, and they grow nervous faster.

A mid-October episode illustrated the new pressure point. Bonds issued by CoreWeave, a company that rents out AI chips to tech giants, sank to yields around 15%, up from roughly 10% this summer. The Philadelphia Semiconductor Index fell 11.4% in the week ending October 16.

The checklist sorts 13 signals into three stages: Froth, Cracks, and Confirmation. In the Froth stage, all three signals are flashing. Debt-funded spending is flashing because Oracle posted negative $5.4 billion of free cash flow last quarter on $28.5 billion of spending. Vendor financing is flashing because Broadcom has agreed to lend AI lab Anthropic up to $42 billion to help pay for a $125 billion lease of Broadcom-designed chips. Record stock sales are flashing, with Anthropic preparing an IPO valued at $1.8 trillion to $2 trillion.

In the Cracks stage, two signals have hit tripwires and two are flashing. AI credit spreads are triggered, with the cost to insure CoreWeave's debt against default topping 10 percentage points a year. On October 15, Taiwan Semiconductor reported record profit up 68.7% and said AI demand still exceeds supply, but its U.S. shares fell anyway. Revenue still lags spending by a wide margin, roughly $750 billion of annual spending against much smaller AI revenue. Token prices, the units AI services charge by, hit a record low in September.

The checklist uses a Peak Watch Score that counts how many signals are lit. It updates weekly. The checklist is not yet a sell signal, but it has moved beyond a clear green light.