Nasdaq Notes: A 1.88% Pullback

On October 6 the Nasdaq closed at 27,599.79, a record close, after four straight up weeks. Eight days later it stood at 26,692.87.

What happened in between looks ordinary: down 0.15% on October 13 to 27,204.38, then down 1.88% on October 14, its biggest one-day drop since July 23. Three sessions took off 2.46%, leaving it 3.29% below the high and still up 14.85% for the year.

The numbers aren't frightening. The way it fell is worth a closer look.

Why chips fell first

AI infrastructure and semiconductors led the decline. That's no surprise, but the reason is more specific than "they'd run the most."

A supplier's revenue is, at bottom, Big Tech's capex. More precisely, its growth depends on the increase in capex, not the level. If next year's capex merely stops growing and holds at this year's level, that's just "no longer adding" for Big Tech. For the suppliers, revenue growth drops to roughly zero.

So the moment the market starts doubting how long AI spending can last, this layer is always sold first. It trades on the second derivative of capex, and any hint of slowing gets amplified there.

Three fuses, really one

The tape on October 14: VIX up 10.48% to 16.87, the 10-year Treasury yield at 5.19%, WTI crude at $95.08, gold at $4,119.50, bitcoin at $81,704.

The usual explanation lists three fuses: a hot CPI print that pushed back rate cuts, unease over AI capex and debt, and oil adding fuel. Taken separately they look like three stories. Together, they're connected.

With oil above $95, CPI struggles to cool; if CPI doesn't cool, long-term rates don't come down. A 5.19% 10-year means the premium for owning stocks over Treasuries gets squeezed thin. Richly valued growth stocks are held up by "higher profits later," and higher rates make those distant profits worth less today.

The more important link runs through AI capex itself. More and more of this round of spending is funded by debt and financing. The higher rates go, the higher the return bar that spending has to clear. So rates and the AI story aren't two things: rising rates directly raise the pass mark for whether AI investment is worth it.

A pullback, not a breakdown

That's the pressure. Here's my read: this is a fast pullback from record highs, not a broken trend.

Up 14.85% for the year and 3 points off the top, a drawdown this size after four up weeks is normal pacing. The VIX jump looks big, but 16.87 is still a low level: the market is nervous, nowhere near panic. The hourly RSI hit 22.93; oversold readings like that usually line up with at least a pause, and shorting here doesn't offer good odds.

In other words, the short-term price action alone doesn't show that anything has broken.

What hangs overhead

What really needs watching is whether the AI valuation story itself is being repriced.

There were three signals this week. The IMF warned that a sharp drop in AI stocks could threaten global financial stability. Broadcom's CEO compared some AI investment to gambling. News about CXMT weighed on Asian chip stocks.

The IMF's warning isn't new; flagging risk is its job. Broadcom's comment carries different weight. It sits on the receiving end of AI capex and stands to take a share of every extra dollar spent. When the one collecting the money says some people are gambling, it means people inside the market are starting to separate spending that earns a return from spending that's just keeping up.

Meanwhile Big Tech's budgets keep climbing. Microsoft plans to take data-center capacity from about 12GW to more than 38GW, and Meta keeps pouring money into frontier models. On one side, spending is accelerating; on the other, insiders are reassessing whether it can last. Valuation multiples sit squeezed between the two.

What I'm watching next

One: capex guidance in Big Tech's Q3 reports, which start landing at the end of the month. I'll look at the growth rate, not just the dollar amount. If guidance keeps rising, the companies still see returns themselves, and this is most likely a shakeout. If growth starts to slow, the supplier layer will feel it first, multiple compression has room to run, and that's more than a single pullback.

Two: the 26,000 area near the September 18 low, the first place to see whether buyers step in. If it holds, the high-level range is intact; if not, I'll re-examine my read.

Three: rates and oil. They don't set the direction for AI stocks, but they set how high the return bar is. ASML just beat on AI demand, which says orders haven't broken yet; I'll read that against the capex guidance.

I don't call levels and I don't guess tomorrow's colour. All I can write down is what I'm watching and why.

This is just my own reading, not investment advice.