July 29, 2026 | The AI Boom Just Got a Bill in the Mail

The Fed didn’t hike rates last week. That was supposed to be the good news.

But while everyone was celebrating the Fed staying put, something more important was happening in the bond market — and it’s the real reason tech stocks got hammered.

Here’s the simple version: long-term interest rates are climbing, and that’s a problem for AI.

Why This Matters to You

The AI trade has long been fixated on demand indicators, viewing massive capital expenditure as validation of the boom. However, market attention ignored a critical follow-up: Who ultimately funds these capital investments, and what is the cost of capital behind them? That question just got a lot more expensive to ignore. While short-term rates held steady, the 30-year Treasury yield pushed back toward levels we haven’t seen in almost 20 years. Oil prices climbing, inflation fears creeping back in — the bond market did what the Fed wouldn’t.

Translation: even though the Fed held rates flat, borrowing got more expensive anyway.

US 30-Year Treasury Yield — back above 5.2%, its highest level in nearly two decades, even as the Fed held its overnight rate steady.

Why AI Companies Care So Much About This

Meta just told investors it plans to spend up to $145 billion in 2026 building AI infrastructure. That’s not pocket change — and it’s not all cash sitting in the bank. A lot of this buildout is financed: debt, leases, credit from suppliers.

It’s a good story until the interest bill shows up.

The Spark: Korea

The selling actually started in Korea. SK Hynix — a major chipmaker — put up genuinely strong results. In a normal market, that’s a ‘buy the dip’ headline.

Not this time. Revenue came in just a touch below sky-high expectations, and the reaction was brutal. It didn’t stay contained to one stock, either — the entire KOSPI index rolled over, sliding out of its uptrend and into a clean downtrend that’s still running.

The lesson: when a stock — or a whole market — is priced for perfection, ‘very good’ isn’t good enough anymore.

KOSPI Index — Korea’s benchmark rolled out of its uptrend and into a falling channel the same week SK Hynix’s results disappointed, spreading the selloff well beyond a single stock.

That selling spread fast — into chips, into AI-adjacent names, into the broader market globally.

SOXX Semiconductor ETF — the clean uptrend (green channel) broke the same week Korea’s chip selloff hit, and price has been sliding in a new downtrend (red channel) ever since.

And here’s the catch-22: the more confident these companies get about AI demand, the more they spend to keep up — which means the more they need to borrow — which means the more exposed they are when borrowing costs rise.

What Changes From Here

This doesn’t mean the AI trade is over. It means the market is about to get pickier.

Going forward, expect investors to start separating AI winners into two camps:

• Companies already turning AI demand into real cash flow — these should hold up.
• Companies still burning cash building for a future that hasn’t arrived yet — these get judged much more harshly.

Cheap chips and big spending numbers won’t automatically justify sky-high valuations anymore. The market wants to know the financing terms, not just the growth story.

Bottom line: The AI trade isn’t broken. But the free pass on ‘spend now, profit later’ is over. From here, it’s not just about how big the story is — it’s about who can actually afford to build it.

Martin Straith July 29th, 2026

Posted In: The Trend Letter

Next: