Elevator Down: Did the AI Selloff Crack the Foundation?

Thursday was the worst day for U.S. stocks since April 2025, and the AI corner of the market took the elevator. The bigger question is whether anything underneath it gave way.

Data as of the U.S. market close, 4:00 p.m. ET, Friday, November 6, 2026. Later information belongs in our next update.

If you’ve read my letters for a while, you know the line I keep coming back to: when the market goes up it takes the escalator, and when it goes down it takes the elevator. On Thursday, November 5, the AI floor found out just how fast that elevator moves. The S&P 500 fell 4.4%, the Nasdaq-100 6.4% and chip stocks more than 11%, all in one session [1][2].

So, did the foundation crack?

Not yet, based on what the team has seen. High-yield bonds slipped only about 0.3%, Treasuries rallied and the team found no strain in bank funding markets [2][9][11]. Stocks steadied on Friday after a strong jobs report [2][12]. That doesn’t make me comfortable, though. The cracks the team was watching before Thursday are all still there, which is why our view stays cautious.

This brings me to this month’s update and what I am calling “Elevator Down.”

Who was on the elevator?

Two days after the midterms, in which news organizations projected Democratic control of both chambers of Congress, the post-election party ended before anyone swept up the confetti. The S&P 500 closed at 7,353.73, its steepest drop since the April 2025 tariff announcements [1]. All 11 sectors fell, and technology dropped 7.3%, its weakest day since March 2020 [3]. On the New York Stock Exchange, 2,344 stocks fell and only 416 rose [3].

And yet, here’s the part that didn’t make the headlines. The equal-weighted S&P 500, which gives every company the same size in the index, fell 2.9%, less than the index itself, because the largest technology companies did most of the damage [2]. I’ve been beating this drum since early 2024: when a handful of companies carry the index up the escalator, they can carry it down the elevator just as fast. Thursday was that day.

So why Thursday? Senior Democrats called for limits on new data centers, and some winners of state races had campaigned on pauses, which raised doubts about the pace of the AI build-out [5]. Then a large chip designer gave a weak outlook [6]. If that wasn’t enough, a widely followed bank strategist described hedge funds rushing to lock in their 2026 gains in crowded AI positions [7]. None of these explanations is proven yet, and calls for limits aren’t rules. The team will treat them as policy only once something is formally filed.

The fear traveled overnight. Korea’s main index fell more than 8% and briefly hit a circuit breaker, Japan’s Nikkei fell about 6% and Taiwan more than 8% [8]. One strategist quoted by Reuters saw “clearly an element of panic in the selling” [8].

The VIX, Wall Street’s fear gauge, closed at 28.18 on Thursday, up from 16.61 the day before [4]. Its three-month cousin closed lower, at 25.55 [4]. In plain English, investors were paying more for protection over the next few weeks than over the next few months. That’s what acute fear looks like, and it’s different from a settled belief that the economy is getting worse.

Did the foundation move?

Think of a building in a storm. The penthouse can sway, but what you really care about is whether the foundation shifts. In markets, the foundation is credit and bank funding. That’s where trouble in stocks turns into trouble for borrowers, then banks, and eventually jobs.

So let’s look under the hood. High-yield bond spreads, the extra yield investors demand to lend to riskier companies, widened from 3.32% to 3.48% [9]. That’s a real move, and still a small one next to a 4.4% drop in stocks. As a point of reference, in the April 2025 selloff a similar fall in stocks came with a jump in spreads more than three times as large [9]. The 10-year Treasury yield fell to 4.84% from 4.90% as money looked for safety [10]. The New York Fed’s standing repo facility, the backstop for banks that need cash overnight, saw no take-up at all on Thursday [11]. Banks and financial stocks fell, but less than the market [2].

Friday brought a little calm. The S&P 500 rose 0.3%, the Nasdaq-100 1.4% and chip stocks regained 3.6%, while the equal-weighted index slipped 0.4% [2]. For the week, the S&P 500 lost 2.7% and sits about 6% below its October record, and the Nasdaq-100 is about 14% below its high [2].

So far, so good… for the foundation. A building that makes it through one storm hasn’t proven much about the next one.

The Canary reading, in plain words

Our Canary Dashboard reads 6.52 out of 10 as of 2 a.m. ET on November 6, in its Elevated band, and it barely moved during the week. It brings the team’s risk ratings for each part of the economy and markets, from credit and banking to energy, housing and geopolitics, into one number on a scale of 1 to 10, where 10 is the most concern.

Two cautions. First, that reading was taken before Thursday’s volatility and credit spread data reached the team’s system, and the team estimates that including them lifts it only slightly. Second, it is not comparable with the 7.8 we published on September 24, 2026. That number came from our earlier framework, which combined a smaller set of measures in a different way. Same name, different ruler.

The highest readings remain in geopolitics and energy, followed by valuation, market breadth, interest rates, private credit and housing. The ratings for AI investment and breadth rose after Thursday, and the valuation rating eased simply because prices fell. Company credit and the job market are the calmest areas, and they are the main reason the reading isn’t higher. Notice where those two sit… they’re the foundation.

The team’s overall view remains cautious, with risk seen building at every horizon from three months to a year and a half, most at around six months. Early in the week the outside research we follow leaned mildly constructive. By Friday it leaned mildly cautious. The gap between us narrowed, though not always for the same reasons.

The cracks that were there before Thursday

Money is still expensive. The Federal Reserve held its target range at 3.75% to 4.00% on October 28, with two members preferring an increase [13]. Friday’s jobs report showed 165,000 jobs added in October and unemployment at 4.1%, and traders raised the odds of a December rate increase to about 73% [12]. Good news for workers can still mean higher borrowing costs for everyone else. Higher for longer, anyone?

The AI build-out has also become a balance-sheet story. In the third quarter one large cloud provider spent $51.4 billion on property and equipment against $42.3 billion of operating cash flow, and another spent $77.3 billion against $42.4 billion [14][15]. Then on November 3, a highly rated platform company sold $15 billion of bonds, with its 30-year notes paying 7.50% [16]. When you spend more than you bring in, you borrow the difference. That works fine until the people lending to you start asking harder questions. Thursday was the first real test of that borrowing against a falling stock market.

Then there’s the world outside our borders. Brent crude traded near $99 a barrel on Friday morning in Asia, heading for about a 12% weekly loss, as Reuters reported that crude shipments through the Strait of Hormuz had recovered, even with at least five vessel incidents in the Gulf since November 1 [18]. Late on Thursday, the Commerce Department extended part of the U.S.-China export-control relief to January 10, 2027 [17]. A separate suspension of U.S. port fees on Chinese-linked ships isn’t covered by that rule and can still lapse on paper on November 9 [17].

But maybe the optimists are right?

I’m not a glass half empty person, so let’s give the other side its due. The economy grew at a 3.4% annual rate in the third quarter [19]. October’s hiring beat forecasts by a wide margin, with upward revisions to earlier months [12]. Credit and bank funding absorbed a severe stock market shock with little strain [9][11]. If Thursday was mostly hedge funds trimming crowded positions, it may have cleared out some excess without hurting the economy, and Friday’s bounce in chip stocks fits that story [2][7]. A market where 59% of S&P 500 companies still trade above their 200-day averages isn’t a market in broad retreat [20].

The optimists may be right. The honest answer is that being early and being wrong can look exactly the same for a long time. That’s why the team describes conditions and doesn’t try to call the top.

It’s not the market, it’s your decisions

Thursday and Friday showed how quickly the story can flip. A 4.4% drop invites the urge to sell everything [1]. A bounce the next morning invites the urge to jump right back in. Neither one is a plan.

The more useful question is which floor your own money lives on. If you’re a founder or an executive in technology, Thursday was a reminder that your company, your equity compensation and your investments can all end up riding the same elevator. Business risk shouldn’t become portfolio risk.

What we’re watching next

Credit is the first test: whether high-yield spreads keep widening once Thursday’s move settles, and whether new AI-related bond deals still find buyers. The team is also watching the November 9 port-fee deadline, October inflation data on November 10, a major AI chipmaker’s results on November 17, and whether any data-center limits move from statements to formal rules [5][17].

And here is the evidence that would ease our caution: credit spreads falling back, inflation cooling enough to take a December rate increase off the table, Gulf shipping returning to normal, and AI spending paid for out of cash flow instead of borrowing. If those show up, the Canary Dashboard should move lower, and we’ll say so.

Bottom line: the elevator went down and the foundation held… this time. Only time will tell whether Thursday was a storm passing through or the first crack the team has been watching for.

Sources

[DISCLOSURES: to be added by the CCO before publication]