Wall Street spent the week watching inflation. The bond market saw a bigger shark.

Two inflation reports came in hotter than expected. That is normally the signal for bond investors to sell.

Instead, they bought long-dated Treasuries at the fastest pace in a year and a half.

The iShares 20+ Year Treasury Bond ETF (NASDAQ:TLT) rose 2.55% this week to $79.97 by Friday morning.

It is on pace for its best week since March 2025. Just nine days ago, on Oct. 7, the fund traded at $76.43, the lowest price in its history going back to 2002.

So why buy bonds when inflation is heating up? The answer is artificial intelligence.

Hot Inflation Should Have Sunk Long-Term Treasuries

On Wednesday, the Bureau of Labor Statistics said consumer prices rose 0.6% in September. That pushed annual inflation to 3.7% from 3.4%, above the 3.6% economists expected.

Core inflation, which strips out food and energy, rose to 2.6% annually, still topping the Fed's 2% target.

Thursday's producer price report was hotter still. Headline producer inflation soared to 6.3% year over year, the highest print since December 2022.

On a monthly basis, prices jumped 1.3%, more than double economists' 0.5% forecast.

The first reaction followed the script. The 10-year Treasury yield briefly climbed above 5.32% after the consumer price report.

Then it reversed.

By Friday, the 10-year yield was at 5.10%, down 14 basis points from last Friday. The 30-year yield fell 16 basis points to 5.44%, after touching 5.73% last week, its highest level since 2001.

Bond prices move opposite to yields, which is why TLT rises when long-term yields fall.

The Bond Market Is Pricing a Slower AI Boom

The selling that mattered this week hit AI debt, not Treasuries.

On Thursday, CoreWeave Inc.'s (NASDAQ:CRWV) bonds due 2032 fell below 80 cents on the dollar, with yields above 15%.

CoreWeave's five-year credit default swaps rose to 1,194.5 basis points, a record intraday high, according to ICE Data Services data reported by Bloomberg.

At that level, the swaps imply a more than 60% chance of default within five years.

On Wednesday, Gartner, one of the world's largest technology research and advisory firms, estimated that AI data center operators need $8.5 trillion in cumulative revenue through 2030 to cover their spending, financing costs and a 12% return on invested capital. It forecasts no more than $3.6 trillion.

As a result, Gartner said more than $3.5 trillion of planned AI infrastructure spending is exposed to write-downs.

The chain works in steps:

• The AI buildout has become a major engine of U.S. growth.
• If lenders demand more to fund it, some projects get delayed.
• Less investment means slower growth.
• Slower growth means lower long-term yields.

"When expected activity in AI declines, yields decline," Dennis DeBusschere, analyst at 22V Research, said in a note Thursday.

Tech stocks felt it.

The Invesco QQQ Trust (NASDAQ:QQQ), which tracks the Nasdaq 100, is down 3.1% this week.

The Long End Is Doing the Work, Not the Fed

The rally was concentrated in long-dated bonds. The 2-year yield, which tracks expectations for Federal Reserve policy, slipped only 2 basis points to 4.74%.

Meanwhile, traders have stepped back from an immediate rate hike.

Fed funds futures price in a 14.9% chance that the Fed will raise rates at its Oct. 28 meeting, down from 19.5% a week ago, according to CME FedWatch.

In other words, the bond rally bets on slower growth, not cheaper money.

One Good Week After a Long Slide

Not everyone is ready to call the AI spending cycle over.

"We are nowhere near ready to write off the AI investment boom," Ed Yardeni said in a note.

He noted that Amazon.com Inc., Alphabet Inc., Meta Platforms Inc., Microsoft Corp. and Oracle Corp. are expected to spend a combined $892 billion this year and $1.1 trillion next year.

TLT is still down 8.5% since January.

One strong week does not reverse a slide in long-term bonds that began in the fall of 2024. For now, it shows bond investors fear an AI slowdown more than inflation.