It was the week Wall Street wavered on America's Teflon economy.

A top chipmaker's chief executive likened some AI investors' behavior to gambling. Reports of Chinese advances in semiconductors added fuel to a rout in chip stocks. A hot inflation report sent the bond market on a roller coaster and, along the way, one of the most prominent AI cloud companies had a meltdown in both the equity and debt markets.

A broad tech selloff Friday punctuated the market's worst week since June as the Nasdaq composite tumbled 4.5%.

Hovering over all of it were fears among investors that the higher borrowing costs in the U.S. might slow the historic investments in artificial-intelligence infrastructure that have become an all-important driver of economic growth. That, in turn, led to a flight back to government bonds, ending a bruising selloff that had driven Treasury yields to their highest levels in more than two decades.

As bonds rallied, the yield on the 10-year Treasury note fell 0.146 percentage point to 5.097%, according to Tradeweb -- its biggest weekly decline since mid-February, before the start of the Iran war.

The week's events bolstered the arguments of investors like Ben Inker, co-head of asset allocation at Boston-based money manager GMO, who has been a vocal bear of the AI trade.

He points to prior periods of heavy investment such as the railroad boom and mania around dot-com stocks, which led to market busts. To him, AI is no different from those frenzies.

"If AI spending does not continue to grow, that gets tough for aggregate demand," Inker said. "It's likely that if AI really slows down, it will be really bad on the economy."

One recent warning sign? Rising borrowing costs for AI companies, along with a looming deluge in share issuance from companies including Anthropic and OpenAI. "You've got to bribe the lenders that much more to take more data center risk," Inker said.

The fears this week trickled through far-reaching corners of the market.

Riskier and more heavily indebted companies such as the cloud computing upstart CoreWeave took the brunt of the blow, with CoreWeave shares sliding 12% on Friday and 25% for the week. But the selloff swept through the wider tech sector, bringing down everyone from giant hyperscalers Amazon.com and Alphabet to highflying chip stocks Micron and Advanced Micro Devices.

The PHLX semiconductor index, packed with AI heavyweights like Nvidia and Broadcom, slid 3.8% on Friday to mark its longest losing streak since 2008. Dell Technologies plunged 22% this week, its worst weekly performance on record.

The swoon in stocks highlighted the close linkages between the economy, markets and a relatively small group of technology highfliers. There are roughly 45 S&P 500 companies deemed as having material exposure to artificial intelligence. Together, those companies have made up around 39% of the index's 14.1% return through Oct. 11, according to S&P Dow Jones Indices.

"AI is not only the economy, but it has become the equity market and a dominant force in the bond market too," Barclays strategists wrote in a recent note to clients.

The S&P 500 fell 1.1% on Friday, bringing its weekly loss to 1.9%. The Dow Jones Industrial Average, which carries relatively little exposure to tech, edged lower by 0.3%, or 175 points. The blue-chip index eked out a gain of 0.7% for the week.

Many see the AI trade, once a monolithic bet on chip stocks and hyperscalers, splintering. "Picks-and-shovels" companies making AI hardware are better-positioned than those relying on "cheap borrowed money to expand," said Michael Landsberg, chief investment officer at Landsberg Bennett Private Wealth Management.

"This is not a riskless trade. There's definitely going to be winners and losers," said Landsberg.

Warning signs started to flash in the trading of bonds tied to data-center companies this week, many of which fell sharply. The losses started with the riskiest and most indebted companies, like CoreWeave, but by Friday were showing up in data-center bonds with investment-grade ratings. Analysts said a Thursday announcement that Apple and Meta would team up on new AI models that run locally on a laptop, rather than at data centers, added to concerns.

For months, U.S. Treasury yields had been propelled higher by a mixture of pessimism about inflation and optimism about economic growth. Investors have thought that the Federal Reserve would raise rates further to fight inflation but that it would take much higher rates to seriously curb economic growth powered by an AI build-out seemingly unaffected by changes in borrowing costs.

Investors still expect the Fed to raise rates again -- if not this month, then soon thereafter.

Federal Reserve Chairman Kevin Warsh said little about his thinking on rates in a public appearance Friday morning local time in Bangkok. Investors took that to mean that the central bank likely won't raise rates at its Oct. 27-28 meeting given that other officials have previously signaled that a pause was likely. But interest-rate futures indicate that traders still see a roughly 85% chance that the Fed will raise rates by the end of the year, with this week's hot inflation data only cementing those bets.

Some debt analysts focused on the tech sector remain skeptical that companies will slow their investments in AI infrastructure just because they are facing higher borrowing costs.

For one thing, they say, companies have already funded much of the spending that they plan to do over the next year or so. Demand for AI computing power also remains so high that even riskier companies have huge incentives to keep investing and finance their expenditures through whatever means possible.

"In order to really create a slowdown, you need to have the demand slow down," said Rich Gross, a senior fixed-income analyst at Columbia Threadneedle Investments. He added, though, that rising rates could curb AI investment in other ways, such as by hurting consumers and generally slowing the economy.