The Nasdaq Fell 5% In A Day. What Happened The Last 21 Times?
Follow the whole story: every piece in this series is on our AI storyline page.
The past can’t tell you what Friday does. It can tell you something more useful: whether a plan that holds through days like this has paid off before. We’ll come back to that at the end.
What Does Everyone Already Know?
Thursday was a rough one. Using the big index funds as stand-ins, the S&P 500 (SPY) fell 4.4% to $733.00 and the Nasdaq-100 (QQQ) 6.5%. The Dow fund and the small-company fund fell less than the Nasdaq, while gold (GLD) and long Treasuries (TLT) edged up.
It was the AI trade unwinding: the chip makers and the companies building the power and data centers behind AI fell hardest. The worry has been building for weeks. Power is the bottleneck, opposition to new data centers is growing town by town, and the new Congress can lean on both; on Wednesday, Meta’s president told a Wall Street Journal conference that the midterm results will make building data centers harder. We flagged power as the AI trade’s real risk on Nov 2, in Nvidia Stock’s Real Risk Isn’t China. It’s Power.
The damage came from the very top of the market. Ten of the biggest names we track lost roughly $1.5 trillion of market value between them in one session:
| Company | Thursday | Value lost | Below 52-wk high |
|---|---|---|---|
| Nvidia | −9.3% | $448B | −24% |
| Broadcom | −10.4% | $162B | −39% |
| Micron | −17.5% | $157B | −46% |
| Microsoft | −3.6% | $148B | −4% |
| AMD | −17.3% | $138B | −38% |
| Tesla | −10.8% | $134B | −30% |
| Apple | −1.5% | $78B | −2% |
| Alphabet | −2.2% | $78B | −30% |
| Meta | −4.0% | $64B | −22% |
| Amazon | −2.2% | $63B | −8% |
Nvidia alone lost about as much value as Apple, Alphabet, Meta, Amazon and Microsoft combined.
The chip makers lost the most, and the companies that buy their chips lost far less; we take that split apart in Is The AI Trade Officially In Trouble? So here’s the question nearly everyone has after Thursday: is the market overreacting? Nobody can tell you what Friday does. But we can tell you what happened every other time it looked like this.
How Often Has This Happened Before?
We went through every trading day in the Trefis price cache from December 31, 2002 to November 4, 2026 and kept the days when the S&P 500 fell at least 3.5% and the Nasdaq Composite fell at least 5%. There were 21. Thursday, with the S&P 500 down 4.4% and the Nasdaq Composite down 5.5%, is the 22nd.
| When | Days | Higher 1 week later | Higher 1 year later |
|---|---|---|---|
| Fall 2008 | 10 | 5 of 10 | 9 of 10 |
| Jan 2009 | 1 | 1 of 1 | 1 of 1 |
| Aug 2011 | 3 | 2 of 3 | 3 of 3 |
| 2020 | 4 | 1 of 4 | 4 of 4 |
| Sep 2022 | 1 | 0 of 1 | 1 of 1 |
| Apr 2025 | 2 | 1 of 2 | 2 of 2 |
Days like Thursday travel in packs. Eleven of the 21 came in one stretch, the 2008-09 crisis.
A 5% Nasdaq day is rarely alone. In the fall of 2008 there were ten of them in about nine weeks. In March 2020 there were three in eight days. One of the 21 landed on November 5, 2008, the day after a presidential election. Thursday came two days after this year’s midterms.
What Happened Next?
Here’s the S&P 500 after all 21 days, price only, no dividends:
| Later | Higher | Median | Worst |
|---|---|---|---|
| 1 week | 10 of 21 | −1.2% | −13.1% |
| 1 month | 14 of 21 | +4.6% | −15.0% |
| 3 months | 12 of 21 | +2.8% | −21.4% |
| 12 months | 20 of 21 | +27.0% | −4.1% |
| 5 years | 18 of 18 | +100% | +53.2% |
The next week and the next three months were close to a coin flip. The next year and the next five were not.
A week later the range ran from a further double-digit fall (after March 9, 2020) to a double-digit gain (after November 20, 2008). That’s noise, not a pattern you can trade. Three months later the worst case came after September 29, 2008, with more of that crisis still ahead.
Then the picture changes. A year later the index was higher 20 times out of 21, with a median gain of 27%. The one miss was that same September 29, 2008 day, a bit lower a year on. Five years later, for the 18 days old enough to measure, it was higher every time, roughly doubling at the median.
See more lines: the fine print on these numbers
Price only: dividends would add to every line. The 2022 and 2025 days aren’t five years old yet, so the five-year row covers 18 days. The days overlap (ten in fall 2008), so this is really five or six episodes, not 21 independent tests. The window starts on December 31, 2002, after the 2000-2002 dot-com fall, so that bear market is not in this data set. Thursday’s figures are the SPY and QQQ funds; the precedents use the S&P 500 and Nasdaq Composite indexes.
How Low Did It Go In The Worst Episodes?
A median hides the hard part, so here it is. In each of the last three bear markets, here’s how far the funds and the market’s biggest AI name fell from peak to bottom, and how many months it took to get back to the old high from the bottom, on dividend-adjusted daily prices:
| Fund or stock | 2008-09 | 2020 | 2022 |
|---|---|---|---|
| SPY | −55%, 41 mo | −34%, 5 mo | −24%, 14 mo |
| QQQ | −53%, 25 mo | −29%, 3 mo | −35%, 13 mo |
| SMH (chips) | −62%, 39 mo | −34%, 3 mo | −45%, 9 mo |
| Nvidia | −85%, 89 mo | −38%, 2 mo | −66%, 7 mo |
Every one of these came all the way back. Some took months, and 2008 took years.
Thursday’s drop is small next to any of those rows. Nvidia ended the day about a quarter below its high, against a two-thirds fall in 2022. That doesn’t mean it can’t go lower. It means the history in our first two tables already includes falls far worse than Thursday, and the year-later numbers still came out the way they did.
So Is The Market Overreacting?
Here’s the honest answer: the data can’t say about this week, and neither can anyone else. What it can say is which question has an answer. “Will it bounce Friday?” had a coin-flip answer 21 times. “Will I be glad I held on a year from now?” had a nearly one-sided answer 21 times. We wrote earlier this week about whether this Nasdaq selloff looks more like 2000 or 2018; that’s the piece to read if your worry is the AI trade itself rather than the market.
“We live in decades. Why worry about the next ten minutes like a fruit fly?” – Anonymous
What Should You Do This Week?
Make a plan for the next three months, not a trade for Monday. Three steps:
1. Assume there may be more days like this. In 2008 and 2020 they came in clusters. A week after two of the 21 days the index was already 10% or more lower again. Decide now what you’d do if SPY fell another 10%, to about $660. Write it down.
2. Rebalance by rule, not by mood. If your stocks have drifted well below your target share, a pre-set band tells you when to buy back to it. If they haven’t, you do nothing, and that’s a decision too.
3. Check your chip exposure. Thursday’s biggest falls were in chips. An S&P 500 fund already holds Nvidia, Broadcom, AMD and Micron; Nvidia and Broadcom alone are about a tenth of SPY. Count them before adding more.
The risk: the worst three-month case in the data was another 21% fall. A plan doesn’t prevent that; it means you’ve already decided what you do inside it. What it replaces: selling after a 5% day, or buying everything at once because “it always bounces.” It didn’t always bounce within a week.
What to watch: whether more 5% days follow in a cluster, and Nvidia’s quarterly report later this month. Both are reasons to check your plan, not to trade.
What’s The Bigger Idea?
We call it Plan, don’t react. The scariest day in a year is rarely the one that decides the year. Of the 21 days like Thursday since 2002, the one that hurt most a year later came in the thick of the 2008 crisis, and even it was 53% higher five years on. The short run is loud. The long run, in this data, was kinder.
History says Thursday’s drop told you little about next week and a lot about how to plan. After the 21 earlier days like it, the S&P 500 was higher a week later only 10 times and three months later 12 times, but a year later 20 times, median +27% (price only). The bear markets around them took SPY 5 to 41 months to recover. Plan for more rough days, rebalance by rule, and judge this by where you’ll be in a year, not on Friday.
What Should You Ask Next?
- Will stocks rebound today, or do you have a plan? The one-page plan, step by step.
- Is the AI trade officially in trouble? The chip makers against the companies that pay them, metric by metric.
Follow this story with us: if more days like Thursday come, we’ll add them to the table and show what changed.
What Would Your Plan Do After Another 10% Drop?
The next three months may be rough; your plan doesn’t have to be. If you have $1 million or more invested, get a second opinion from a Trefis expert partner advisor in a free 15-minute call, or let Invest With Trefis run the plan for you.
For questions about this article or underlying research, email help@trefis.com.