The other side of the machine
S&P 500 6,832.76 · −108.71 · −1.57%
For two years the artificial-intelligence build had been a reason to own almost anything. On 12 February the market spent a day considering the same build from the other end — as a thing that arrives at a business and removes the need for most of it — and it did not enjoy the exercise. The index made its high of 6,973.22 ten minutes after the open, gave up a hundred and eighteen points by twenty past eleven, spent five hours going nowhere in particular, and then broke in the last half hour to finish at 6,832.76, down 1.57 percent, with the low of the day printed inside the final ten minutes. Utilities rose. Packaged food rose. That is what this kind of fear looks like while it is still polite: not a run for the exit, but a quiet, orderly move into the two industries that no software has ever been able to explain away.
The day's protagonist
Cisco Systems CSCO
Move: −12.32% to 75.00 — the fear arrives with a number attached
Cisco makes the switches and routers that the data centres are built out of. If the artificial-intelligence build were an unambiguous good for the companies that supply it, Cisco is the last place you would expect the doubt to land. It guided the current quarter down anyway, and the stock opened at 79.61 — a gap of nearly six dollars below Wednesday's close, before a single share changed hands in the session — traded down to 74.085, and finished at exactly 75.00, off 12.32 percent, on twice its normal volume. It was the worst thing in the Dow and it dragged the average down 669 points. What made the day frightening was not the guidance, which was one company's quarter. It was that the guidance was the first hard number anybody could attach to a soft, widely held, entirely unquantified worry: that the build could be enormous and real and still not pay the people building it. Apple fell exactly five percent the same afternoon on no news at all. Money went into utilities, up 1.5 percent, and consumer staples, up 0.9 — the two sectors whose businesses are the hardest to describe as software problems.
The deeper account
The selling came in two pieces with a long flat stretch between them, which is a useful thing to look at. The first piece was the morning: from 6,973.22 at twenty to ten down to 6,854.94 by twenty past eleven, a hundred and eighteen points in under two hours, which is a market reacting. Then nothing — five hours in a forty-point band, up a little, down a little, the shape of a market that has priced the news and is waiting to see whether anyone disagrees. The second piece was the last thirty-five minutes, when it went out on its low. Afternoon lows are the tell. A market that gaps down and recovers has absorbed something; a market that spends the day steady and then breaks at the bell has decided something.
The thing it decided was not about Cisco. Cisco's quarter was the occasion. The subject was a proposition that had been floating around, unpriced, for most of a year: that the artificial-intelligence build might be simultaneously the largest capital programme in modern corporate history and a net destroyer of the earnings of most of the companies that are not building it. Every previous version of that worry had been abstract. This one came with a number and a share price, from a company that sells the shovels, which is why a single set of quarterly guidance took two percent out of the Nasdaq.
What the money did next is the part worth recording. It did not leave. It moved into utilities and consumer staples — power and groceries — which rose while everything else fell. That is not panic; panic sells everything and asks later. It is a specific, considered proposition about which businesses a model can be pointed at, and it is the same proposition, in a much politer form, that the market would spend the rest of February arguing with itself about. Then, on the last day of the month, the war started, and for the next five weeks nobody discussed artificial intelligence at all.
The artifact
The ledger — what actually happened
| Measure | Close | Change | Marginalia |
|---|---|---|---|
| S&P 500 | 6,832.76 | −108.71 · −1.57% | high 6,973.22 at 9:40 · down to 6,854.94 by 11:20 · five hours sideways · low 6,824.04 in the final ten minutes |
| Nasdaq Composite | 22,597.15 | −2.03% | down 469.32 points; the Dow fell 669.42, or 1.34 percent, led lower by Cisco |
| Cisco Systems | 75.00 | −12.32% | guided the current quarter down · gapped open at 79.61, low 74.085, closed on a round number · 68.1 million shares against 34.5 million the day before |
| Apple | 261.73 | −5.0% | five percent to the second decimal place, on no company news — the sector being repriced rather than the company |
| The sold | tech −2.6% | financials −2.0%, communication services −1.8%. The complex that had led the market for two years led it down | |
| The bought | utilities +1.5% | consumer staples +0.9%. Power and groceries: the two things nobody has yet claimed a model can do instead | |
| Six days earlier | 50,115.67 | the Dow's first close above fifty thousand, on 6 February. It does not trade above it again in this batch |
Sources
- Robinhood market-data feed — SPX 10-minute bars and day bar; CSCO, AAPL, INTC and CAT day bars, adjustment none. Pulled 2026-08-28
- Zacks via Nasdaq — "Stock Market News for Feb 13, 2026" (covers the 12 February session: index closes, sector ETFs, Cisco's guidance and move) — https://www.nasdaq.com/articles/stock-market-news-feb-13-2026
- CNBC — "Dow slides 600 points, S&P 500 falls for a third day as AI disruption fears rattle markets" — https://www.cnbc.com/2026/02/11/stock-market-today-live-updates.html