Is SMH A Great Dip Buy At 31% Off?
Today’s big story: Could Nvidia Stock Fall 50%? Follow The Cash
Chip stocks ran hard, then broke. The VanEck Semiconductor ETF, SMH, hit a closing high of $668.91 on June 22. It closed Monday at $461.90, down 5.1% on the day and 31% below that high. Even after that fall, SMH is still up 34% over the past year. The S&P 500 fund SPY is up 11%. This is a fall from a very high place.
So is it a great dip? We answer with every fall this big since 2003. Then we widen the view, because a dip is a 1-year question and AI is a 10-year one.

What Happened After Every Fall This Big?
The record is short but strong. In all four past falls of 30% or more since 2003, SMH was higher a year after the first day it crossed that line. The median gain was 60%.
| First day 30% down | Further fall | 1 year later |
|---|---|---|
| Jul 2004 | −51% | +16% |
| Mar 2020 | −1% | +135% |
| May 2022 | −22% | +14% |
| Apr 2025 | −8% | +105% |
| Nov 2026 (now) | ? | ? |
Source: Trefis, prices adjusted for dividends. Further fall: the lowest close before SMH regained its old high. Before Dec. 20, 2011, the SMH ticker belonged to a different fund, the Semiconductor HOLDRS trust, a fixed basket of chip stocks. The 2004 row uses its prices until VanEck’s ETF took over.
The record holds for smaller falls too. Counting falls of 25% or more, SMH was higher a year later in all 5 past cases, by a median 76%. The median further fall was 8%.
Two of the four big falls healed quickly, in 2020 and 2025. Two were slow. In 2022, SMH first fell another 22% and took over a year to recover. In 2004, the older HOLDRS fund fell another 51%, through the 2008 crash. The ticker didn’t regain its January 2004 high until May 2013, by then as VanEck’s ETF. The 1-year record says buy. The 2004 case says to size it so you could wait nine years.
A fall that takes nine years to heal is easier to sit through when the rest of your money falls less. For every dollar its benchmark lost, the Trefis High Quality (HQ) Portfolio lost about 81 cents on average. See how HQ held up.
Why This Dip May Be Different
Here is the bull case. The next wave of AI, bulls say, will make things. Picture a custom feature film for two on a Friday night, or a 15-minute video coaching your squash swing, made on request. Video takes far more computing than text, and every such request would run on chips.
Bulls also point to the spending. Capital spending at Microsoft, Amazon, Alphabet and Meta reached $615 billion over their last four quarters, up 87% in a year. Nvidia‘s sales were $303 billion over the last four quarters. Analysts expect about $695 billion in the year to January 2028 and $1.2 trillion by the year to January 2031. If the build-out is that early, a 30% fall is a pause, not the end.
The other side, in fairness: those four companies now put 91% of their operating cash into capital spending, and Nvidia’s cash lags its profit. That case is in today’s big story. Nvidia reports third-quarter results after the close today, the next test.
How Do You Buy A Dip Without Betting The House?
Over 10 years, SMH has multiplied 14.8 times, against 4.0 times for SPY. It got there through falls of 30% or more. None of those falls was the end. The 2004 fall, before this window, took nine years to heal. How much to put in, so that either outcome is fine, is worked out with a table in T-Bills At 4.3% Or Nvidia At A 3% Cash Yield Growing 76%?
So is SMH a great dip? Over a year, the record says yes. Size it so you could live through a 2004-style wait.
If you buy in steps, part of your cash will wait on the side for a while.
Waiting To Buy SMH? Why Let Your Cash Sit Idle?
The right price can take months. Meanwhile, the Trefis Income Solution pays a dividend yield about 3x the S&P 500’s, from 50 stocks with at least 10 straight years of dividends. Why not choose a better approach?
For education, not a recommendation to buy or sell.
For questions about this article or underlying research, email help@trefis.com.