Nasdaq −2.6%, Dow Flat (22 October 2026): Why Beats From Meta and Tesla Set Off the Worst Tech Day Since June — and Why Chip Stocks Fell Hardest

The Nasdaq Composite fell 2.60% to 26,008.86 on Thursday 22 October, its steepest drop since early June, while the Dow finished 0.02% higher. The trigger was two earnings beats, not misses: Meta fell 7.0% and Tesla 3.1%, and the Philadelphia semiconductor index dropped 6.7%. The common thread is AI capital spending — a cost on one company's cash-flow statement and revenue on another's income statement, so a doubt about it hits both sides at once.

What happened on 22 October

The closing figures below are from Reuters' market report; the yield and volatility readings are from Yahoo Finance's session coverage.

Volume was 18.79 billion shares against a 20-day average of 17.19 billion. On the NYSE, advancers outnumbered decliners 1.01 to 1; on the Nasdaq, decliners led 1.37 to 1. That breadth matters for the reading: this was a heavy day for a narrow group of very large companies, not a market-wide liquidation.

Paul Nolte of Murphy & Sylvest summed up the split to Reuters: "Investors are starting to realize that there are some holes in the AI story," adding that the indices feel it because they are so heavily weighted toward technology.

Why two beats produced a sell-off

An earnings beat is measured against the income statement — revenue and earnings per share. The concern on 22 October sat on a different statement: cash flow.

Meta's numbers, which this site took apart in Meta's negative free cash flow and the new Class C stock, show the shape. Revenue grew 26%. But capital spending of $34.91 billion exceeded operating cash flow of $32.79 billion, leaving free cash flow at negative $2.12 billion — Reuters described it as Meta's first cash outflow in over a decade. In the same release Meta created a non-voting share class it said was intended to finance growth. The shares rose about 2.5% after hours on the beat, then fell 7.0% in the regular session as the cash-flow reading took over.

Tesla's report pointed the same way from the opposite angle. Adjusted EPS of $0.48 beat a $0.44 estimate and revenue of $28.05 billion was close to the $28.13 billion Bloomberg consensus, per Yahoo Finance. The company lowered 2026 capex — but described higher spending in 2027 and 2028 on Optimus, Cybercab and Terafab. The near-term cash looked better; the longer bill got bigger. The pre-results framing is in the Tesla Q3 preview.

The mechanism is the gap between how long the spending lasts and how long the return takes. Data centres, chips and power contracts are paid for now and depreciated over years. Until revenue from AI services visibly covers that depreciation, the market has to take the return on trust. A quarter where the spending exceeds the cash a business generates makes that trust easier to question. The result was a strong quarter followed by a weak share price, and the two don't contradict each other.

Why chip stocks fell harder than the spenders

The 6.7% fall in the semiconductor index is the sharper part of the story. Most chip companies reported nothing on Thursday. They fell because a hyperscaler's capex budget is their revenue.

The arithmetic of a supplier is about the growth of that budget, not its level. If Meta spends $35 billion a quarter for the next two years, that is a large but flat revenue line for the companies that sell it accelerators, memory and chipmaking tools. Their valuations assume the budgets keep rising. When the market starts asking whether spenders can keep outrunning their cash flow, it is really asking whether the growth rate of capex peaks sooner, and that reaches every supplier at once.

The further up the chain, the stronger the effect. Chipmaking-equipment companies — Lam Research, Applied Materials and KLA were among the session's weakest names, per Yahoo Finance — sell the tools that build new chip capacity. Their orders depend on chipmakers expecting demand to keep growing, so they react to changes in the growth of spending, not just its level. Memory had already been on edge: this site covered the CXMT DRAM sell-off earlier in October.

Why the Dow went up

The three headline indices are built differently, so the same day can look calm or severe depending on which one you watch.

Money that left tech did not leave the market. Procter & Gamble rose 5.6%, Blackstone 8.4% on results, Lockheed Martin 7.0% and Union Pacific 3.6%. For anyone trading index futures, the practical point is that NAS100/NQ and US500/ES carry this AI exposure at very different strengths. On a day like this, the gap between them shows how much of the market's direction comes from a handful of companies — Manulife John Hancock's Matt Miskin told Yahoo Finance that "every index on the planet is basically an AI index."

Not a rates story, but rates are still in the room

The usual explanation for a growth-stock sell-off is rising yields: their value sits in cash flows far in the future, and a higher discount rate shrinks it. Thursday ran the other way. Treasuries rallied, with the 10-year yield falling about 11 basis points to around 5.02% and later dipping below 5%, per Yahoo Finance. When tech falls while yields fall, investors are questioning the cash flows, not the discount rate.

The rates backdrop hasn't gone away. Reuters reported that crude rose to a five-month high after Houthi strikes on Saudi targets, and that markets price roughly a 25 basis-point Fed hike before year-end, with jobless claims near multi-decade lows. An oil-driven inflation scare that pushed yields back up would hit the same stocks through a second channel. On Thursday it didn't happen, and that tells you what kind of sell-off it was.

For currencies the day was a risk-sentiment event more than a rates one. Risk sentiment is one of the five factors the meter scores, and moves like this tend to show up in the defensive currencies first — the yen and the franc — while softer US yields take some support away from the dollar's rate advantage.

What would change the picture

The concern Meta raised is about whether spending is funded and whether it pays back, and the next round of results will be read the same way. Intel has said it reports on 29 October after the close, and Alphabet, Microsoft, Amazon and Apple are expected in the same final week of October. Three outcomes, three different channels:

• Capex guided up, cash flow keeping pace. The ideal case for both sides: suppliers get their growth and spenders show they can afford it. This would ease the doubt Meta raised.
• Capex guided up, funded by debt or equity. Good for suppliers' order books, harder for the spenders' shares — the funding shift that Meta's Class C stock signalled, repeated across the group.
• Capex trimmed or deferred. Comforting for the spenders' free cash flow and difficult for chips and equipment, whose valuations rest on rising budgets. Tesla's move — less in 2026, more in 2027–28 — shows how a "cut" can be a postponement.

None of these settles the bigger question. That comes when AI services revenue shows up in segment disclosures at a scale that covers the depreciation now building on balance sheets. Until then, each capex guide resets how much the market is prepared to take on trust. More on how the site approaches these stories is on the about page.

Educational macro context only — not investment advice.