The Nasdaq Set the Wrong Record

On Wednesday, with the Dow up more than 1,500 points, the President posted that "Your 401(k)s are through the roof."

The next day the Nasdaq lost 1,421 points.

That's the biggest point drop in its 55-year history. The S&P 500 lost 335 points, also the most ever. By Reuters' count, about $3 trillion of market value came off S&P 500 companies in one session.

So yes, Thursday set a record, just the wrong one.

A 5.5% Drop Is the 36th Worst Since 1971

Measured properly, the Nasdaq fell 5.52% on Thursday.

Since the index started in 1971, there have been 35 worse days. The worst was March 16, 2020, when it fell 12.3%. The 1987 crash took 11.35% in a day.

On April 14, 2000, it lost 9.7%, and I remember that one, because I was earning a tech salary that was about to disappear.

Thursday didn't crack the top 30.

Point records are mostly arithmetic. When an index sits at 25,000, a bad day is a thousand points. The record Thursday broke was set on June 5, five months ago, and the one before that was April 2025.

Every one of the Nasdaq's ten biggest point drops has come since 2020, and two of them came in the past six trading days. That tells you more about how high the index climbed than about how hard it fell.

Thursday wasn't nothing. But the headline number was the least interesting thing about it.

The Chip Index Had Its Worst Day Since 2020

The real damage was in one corner of the market.

The Philadelphia Semiconductor Index, which tracks the companies that make the chips the AI buildout runs on, fell 12.25% on Thursday. That was its worst day since March 16, 2020, and it closed 38% below the record it set on June 22.

The Nasdaq-100, where the big tech names live, fell 6.4%, also its worst day since March 2020.

Nvidia fell about 9% and Qualcomm fell 9.5%. CoreWeave, the company that rents out Nvidia's chips, fell about 10%, and two of the companies building data centers for AI lost more than 14% each.

Meanwhile the Dow fell 3.4%, which was only its 19th-worst day of the past decade, and that's a decade that includes COVID.

And unlike October 22, this one had the look of the real thing. (That was the selloff I wrote about in Why Gold Fell With the Tech Stocks, when I told you a real panic doesn't leave half the Nasdaq-100 in the green.)

About 26.9 billion shares changed hands, against an average of 17.7 billion over the previous 20 sessions. The VIX, Wall Street's fear gauge, jumped 70% in a single day, to 28. Since 1990, it's only jumped more than that twice.

So the chips crashed, the fear gauge spiked, and most of the market had an ordinary bad day. Those are three different facts, and I'd keep them separate.

Where the Money Went

The money didn't disappear on Thursday. A lot of it just moved.

Gold rose 4.3% on Wednesday and held its gain Thursday while stocks fell. The main gold-miners fund finished the week up about 14%.

Treasuries rallied too, with the 10-year yield dropping to 4.84%, so somebody was buying bonds while they sold chips. Bitcoin fell about 5% with the Nasdaq, which is what it tends to do on days like that.

And on October 30, the Friday the Nasdaq dropped 750 points, an index of America's highest-dividend stocks hit an all-time high.

Here's the number I find most telling. The extra yield that junk-bond investors demand over Treasuries, which is how the bond market tells you it's scared, went from 3.32% on Wednesday to 3.48% on Thursday. Stocks had a historic day and the bond market barely flinched.

That's good news for now. It also means the bond market hasn't voted yet, and in 1873 it was the bond buyers who decided how the story ended.

The Market Sorted Your Two Piles for You

Two weeks ago, in Big Tech Is Borrowing Like It's 1873, I asked you to sort your stocks into two piles. Companies paying for their growth with cash they earn in one. Companies paying for it with money they borrow in the other.

Thursday did the sorting for you, and it showed the borrowed pile is bigger than it looks.

The borrowed pile holds more than the companies doing the borrowing. Everyone who gets paid by that borrowing sits in it too.

Nvidia's chips, CoreWeave's racks, GE Vernova's turbines and the electricity contracts behind them all get paid out of the same pile of data-center money.

Alphabet alone, as of September 30, had agreed to buy more than $900 billion of goods and services it hadn't received yet, plus about $130 billion of leases on buildings that haven't opened, plus tens of billions backing other people's borrowing.

Call it well over $1 trillion of promises, and none of it shows up as debt on the balance sheet.

By one count of the filings, eight of the biggest AI companies have made close to $4 trillion of these promises between them.

A big share of those promises is future revenue for the chip stocks that fell 12% on Thursday. When investors decided the promises might get stretched out, the companies on the receiving end fell first and hardest, which is what I'd expect.

Why did they decide that on Thursday?

Partly because of Tuesday's vote, in which Democrats won control of Congress and more than 30 local measures to slow or restrict data-center building passed around the country.

Partly because two Fed officials spent Thursday hinting at another rate hike, with inflation at 3.7%.

But mostly, I think, because the bond buyers had been getting more careful for months (orders for Big Tech's bonds thinned from five times the offering in February to under two times by July) and the vote gave them a reason to say so out loud.

1873 Didn't Stop at the Bank Failure

When I wrote about Jay Cooke, I told you the railroad got built anyway and that Cooke got rich again. Here's the part in between.

In 1872, American railroads laid about 7,500 miles of track, and in 1875 they laid about 1,600. By one count, 55 railroads had failed by November 1873 and another 60 within the year, and some 18,000 businesses went under between 1873 and 1875.

The tracks still got used and the country still got built. But the building slowed to a crawl for several years, and the people who'd paid for the first round mostly didn't get to collect on the second.

Nobody's predicting a 65-month depression here. But a buildout doesn't need to stop to hurt the companies that depend on it. It only needs to slow down.

One Move: Look at Thursday and Friday Together

Here's the one thing I'd do this week, and it takes about ten minutes.

Pull up your three biggest holdings and look at two days: what each one did on Thursday, and what it did on Friday.

A stock that fell with the market Thursday and got a good chunk of it back Friday was sold by people who were scared. That's a bad day. The chip index bounced 4% on Friday, for what it's worth.

A stock that fell Thursday and kept falling Friday was sold by people who've changed their minds. That's a different problem, and if it's also in your borrowed-money pile, it's the one I'd look at first.

I wouldn't sell it into this morning's open. If it were me, I'd write down the price where I'd admit I'm wrong, and I'd use the next bounce to get the position back to a size where another Thursday wouldn't rattle me.

We call that number a watch price, and almost every position we own has one. It's a lot easier to pick before the open than at 9:31.

By the way, if your three biggest holdings all did the same thing on Thursday, you own one bet three times. That's worth knowing too.

What Premium Members Get This Week

Everything above is free and yours to keep.

What it can't tell you is what to do with each thing you own this morning. That's the part I save for Premium Members, and this week there's more of it than usual.

After last week, I owe Premium Members a straight answer on every position we own, and that's today's Premium section. Which ones Thursday changed, which ones it didn't, and exactly what we're doing about each, down to the watch price.

Then on Friday morning, they get the new buy alert I told you about last week, with a name, a price and a watch price. The cutoff is Thursday night. Join by then and it's in your inbox when it lands. Join after, and you're getting the name at whatever price it's trading by then.

Here's the record behind it.

14 closed trades. 14 winners. Average gain 83.5%.

Over the same stretches, the S&P 500 did about 34%.

Premium is $35 a month, or $300 for the year when you choose the annual option. That's $25 a month, and you never have to think about it again.

If you've been meaning to join, Thursday night is the cutoff. The review is right below this essay for members, and the new buy alert lands Friday morning.

Double D

🔓 Premium Content Begins Here 🔒