Semiconductors Are in a Bear Market With the S&P 500 Near Its Record. Is That a Warning?

Research note: PHLX Semiconductor index and S&P 500 daily closes May 4, 1994 to October 22, 2026; CRSP industry data 1926 to August 31, 2026; SMH and SPY through October 22, 2026. Frozen October 22, 2026. How we checked it · How we use AI.

• It is the second leg this year. SMH first fell 24.6% below its record in July 2026, recovered to within 5.2% by October 5, 2026, and has fallen 17.4% in the 13 sessions since.
• The S&P 500 usually kept going. A median +19.3% over the next year after the thirteen earlier episodes, against +12.6% for any day near a high.
• The tell was the index’s own high. In eleven of the thirteen cases the S&P 500 closed above its earlier high again, nine of them within three months. In 2000 and 2007 it took years.
• Chips themselves usually had further to fall. The median worst close within a year was another 21.3% lower.
• Chips did not lead past tops. At every S&P 500 bear-market peak since 1973, the chip industry was within 5.0% of its high.

Semiconductors broke down again on October 22, 2026. SMH, the largest chip fund, fell 6.0% and closed 21.8% below the record it set on June 22, 2026, past the 20% line that usually defines a bear market. The S&P 500 lost 1.2% and sat 1.9% below the record it set on October 6, 2026.

My read is that this is a chip problem and not yet a market problem. The split looks alarming, but on 32 years of semiconductor data it has more often come before a good year for the S&P 500 than a bad one. What set the two bad endings apart was visible in the index itself: it did not get back to its high for years. That is the one thing I would watch now.

It is the second break since July. SMH fell 24.6% below its record by July 29, 2026, while SPY barely dipped and then went on to new records. Chips came back to within 5.2% of their record by October 5, 2026 and have given all of it back in 13 sessions.

Thirteen times since 1994

The PHLX Semiconductor index runs back to 1994. On that record, chips closed 20% or more below their 52-week high with the S&P 500 within 5% of its own thirteen times before this year, counting each episode once. The current episode began on July 17, 2026, when the index first crossed the line, and October 22, 2026 is its second leg.

Two things stand out. The first is how ordinary the S&P 500’s path was. It gained +4.2% over the next three months and +19.3% over the next year at the median, better than the +3.3% and +12.6% of an ordinary day near a high. Its worst close within a year was at least 10% lower in only four of the thirteen.

The second is what separated the two bad ones. In eleven episodes the S&P 500 closed above its earlier high again, nine of them within three months and the other two within a year or so. After April 2000 and December 2007, the index had already set its high when chips broke, and it did not see that level again for 7 years and 5 years.

The current episode has already passed that test once. After chips broke on July 17, 2026, the S&P 500 closed at a new high 12 sessions later and went on to its record of October 6, 2026. The second leg starts the clock again.

Three episodes, three endings

Chips themselves usually fell further

For anyone holding chips, the record is less comforting. Within a year of each signal the semiconductor index’s median worst close was another 21.3% lower, and eight of the thirteen episodes went at least 20% deeper. The median change over the following six months was −3.4%. Over a full year the median recovered to +19.2%, but the path ran through lower lows more often than not.

Chips did not lead the market down

The question can also be asked the other way: where were chips when the S&P 500 actually topped? At every bear-market peak since 1973, they were close to their own high. Chips fell with the market, not months ahead of it.

So a semiconductor bear market has not been how tops looked. That does not make today safe; it means the signal people reach for after a day like this one has not been the one that mattered.

How sturdy is the finding?

The 20% and 5% lines are conventions, so I reran the study on 12 combinations, with chips 15% to 30% below their high and the S&P 500 within 3% to 8% of its own. With chips at least 20% down, the S&P 500’s median year beat the near-high baseline in 8 of 9 settings. At 15%, the result was close to, or below, an ordinary year.

A second, survivorship-free record agrees. Ken French’s CRSP chip industry, which moves with the PHLX index with a monthly correlation of 0.91, gives 18 episodes since 1926. The market’s median year afterward was +17.6%, against +12.8% for any day near a high, and it fell into a bear market within a year after five of them (28%), somewhat more often than the 19% of all days near a high. On the semiconductor index the same rate was 15% against 11%. A better median year, then, but not a smaller chance of a bear market.

One version of the setup is unusual for this century. A fall of 15% or more within 15 sessions, with the S&P 500 within 3% of its high, happened six times between 1995 and 2000, never from 2001 to 2025, and two times this year. In the 1990s the S&P 500 rose through the first four. The last two came in 2000, after the index had topped.

What I take from it

Chips breaking down under a calm index has usually been a story about chips. The S&P 500 found other leaders, and the market’s median year was better than average. The exceptions came when the index had already rolled over, which nobody could see on the day, but which showed up within weeks as a failure to get back to the high.

So I would not read October 22, 2026 as the start of a bear market for the S&P 500. I would read it as a bear market in chips, which on this record more often deepened than ended, and as a test of whether the rest of the market can keep the index near its record without them. The July leg passed that test. The October leg has not been tested yet.

What would change my read

Toward the benign endings: the S&P 500 closing above its October 6, 2026 record, which SPY marks at $779.10. In nine of the eleven benign episodes that happened within three months.

Toward 2000 and 2007: the index failing to get back to that high while it slides into a correction. A 10% fall, which SPY would mark near $701, would not settle it alone, because two of the benign episodes also saw one. A 10% fall with no return to the record is what the two bad ones looked like.

How we checked it

• The setup. Chips closing at least 20% below their highest close of the prior 52 weeks while the S&P 500 is within 5% of its own. Each episode counts once: a new one starts only six months after the last began. The rules were set before any outcome was read.
• Chips. The PHLX Semiconductor index since 1994, the longest daily chip index we hold, and SMH for today’s chart. The two weight their members differently, so they sit at different distances from their highs on the same day.
• Outcomes. S&P 500 price changes over the next 3, 6, 12 and 24 months, the worst close within a year, and whether the index fell into a bear market (20% below its 52-week high). The baseline is every day the S&P 500 sat within 5% of its high.
• Robustness. 12 threshold combinations on the semiconductor index, and the same study on Ken French’s CRSP chip industry against the CRSP market with dividends, which includes every listed company of its day.
• Frozen evidence. The tables and charts use the snapshot of October 22, 2026. Only the latest-reading box updates.