Fidelity’s Timmer Says AI Bubble Risk Is Not Stock Valuations, but Debt

Fidelity director of global macro Jurrien Timmer says any AI bubble may sit in debt markets, not stocks. Yields on AI cloud provider CoreWeave’s bonds topped 15% last week, even as equity valuations fall.

Equity valuations are falling as earnings grow, which Timmer finds reassuring. Credit investors, however, are pricing part of the same buildout like a distressed borrower.

Is Credit Flashing a Warning on the AI Bubble?

Timmer has been comparing AI-era price action with the 1998 to 2000 internet boom, he wrote in an X thread.

The big news last week was of course the sharp drawdown in high yield data center bonds and what that might mean for the broader AI buildout. I have been eyeing the price analog between the AI boom today and the internet boom-turned-bust in 1998-2000 and have been comforted by… pic.twitter.com/9YJvlkf7Tb

— Jurrien Timmer (@TimmerFidelity) October 21, 2026

Historically, telecom gave the early warning, as fiber optic networks were overbuilt. That signal came from the debt side, not from dot-com stocks.

He compared a Bloomberg index of junk-rated telecom bonds from that era with CoreWeave’s yield today. The comparison, he said, gives him pause.

Highly rated AI borrowers still look stable, Timmer says, citing the fundamentals of most hyperscalers, the largest cloud companies.

He adds that stress in high-yield AI names could force discipline on the compute buildout. As a result, spending could slow to a more sustainable pace.

Is the Bond Market Wrong About CoreWeave?

CoreWeave began as a crypto miner and now rents out AI computing power. Its 2031 notes yielded 15.78% last Thursday, according to CNBC. Five-year default insurance on its debt briefly cost more than 11% a year.

S&P Global rates CoreWeave B+, below investment grade. Debt rose from $21 billion at the end of 2025 to $35 billion by June 30, its results show. Meanwhile, net interest expense of $640 million took about a quarter of second-quarter revenue.

However, CEO Mike Intrator said in a memo, CNBC reported, that contract backlog more than covers debt obligations. That backlog stood near $104 billion on June 30.

In contrast, Citi’s Tyler Radke, co-head of US software research, attributes the selloff to higher borrowing costs rather than weak demand.

The open question is whether tighter credit trims excess or starves the buildout. Chris White, CEO of bond data firm BondCliQ, says the AI debt glut may lift US government borrowing costs.