The S&P 500 Is Near a Record. Most of Its Stocks Aren't.

mribbeck

16 minutes ago

5 min read

Don Schreiber, Jr.

September 24, 2026

The S&P 500 is knocking on a record while more of its stocks hit new lows than new highs. That combination has appeared only twice in nearly a century: July 1929 and December 1999.

On Monday, September 21, the S&P 500 rose 1.49% to 7,764.70, just 0.67% below its 52-week high. Beneath the surface, 30 of its members hit new 52-week lows while only seven reached new highs (Benzinga).

SentimenTrader's Jason Goepfert found only two other days when the index rallied at least 1% to within 1% of a high while new lows beat new highs: July 23, 1929, and December 21, 1999 (CNBC). Those are not dates anyone wants in their company.

What's happening underneath

A handful of mega-cap technology and AI names are carrying the index while the rest of the market erodes. New lows now outnumber new highs by roughly four or five to one on every venue.

Monday's gains came from communication services, technology and consumer discretionary. Only technology was within 1% of its high; the other two sat 4% and 7% below theirs (CNBC).

The laggards are not fringe names. T-Mobile, Lowe's, McDonald's, Nike, PepsiCo and AutoZone all hit new lows, along with energy and utility stocks (UA.News). CoStar is down 56% this year, and Boston Scientific and Intuit are each down more than 53% (Yahoo Finance).

The macro backdrop adds pressure. The 10-year Treasury yield sits near 4.96% (StreetStats). B. Riley's Art Hogan sees no new highs if the war persists, energy stays high and the Fed keeps hiking.

1998–2000: the closest parallel

The laggards of 1999 became the winners of 2000–2002. The NYSE advance-decline line peaked in spring 1998, yet the S&P 500 kept climbing for almost two more years to its March 2000 peak.

In 1999 the cap-weighted index gained about 21% while the typical stock lagged far behind. Owners of "old economy" companies felt a bear market while headlines celebrated records. The Fed raised rates from 4.75% to 6.5% between June 1999 and May 2000, and oil prices climbed.

The unwind was brutal for the leaders. The Nasdaq fell about 78% from March 2000 to October 2002, and the S&P 500 fell about 49%. Across 2000–2002 the equal-weighted S&P 500 lost roughly 11% cumulatively, against roughly 38% for the cap-weighted index.

1972–73: the Nifty Fifty

An energy shock ended a narrow, high-multiple market, and the parallel to today is hard to ignore. Investors crowded into about 50 "one-decision" growth stocks, such as Polaroid, Avon and Xerox, often at P/E ratios above 40.

Breadth deteriorated through 1971–72 while the S&P 500 pushed to a January 1973 peak. Analysts are now citing January 1973 alongside late 1999 as comparisons (Yahoo Finance).

Inflation and the October 1973 oil embargo broke the spell. The S&P 500 fell about 48% into October 1974, and many Nifty Fifty names fell far more.

2021, 1929 and 2007: more corroboration

2021 was the most recent rehearsal. In November 2021 the Nasdaq Composite hit a record while its new lows outnumbered new highs. Many Nasdaq stocks were already down 50% or more from their highs. The Fed began hiking, and the Nasdaq fell about a third in 2022.

1929 is the extreme bookend. The July 23 signal came about six weeks before the September peak. The Dow then fell about 89% by July 1932. Treat it as a reminder of what narrow markets can become, not as a forecast.

2007 was shorter but clear. The advance-decline line peaked in June 2007 and the S&P 500 in October. The index then fell about 57% into March 2009.

Comparing the episodes

Every major breadth divergence in this study ended with the leaders falling at least a third. The lead time varied widely: about 23 months from breadth peak to index peak in 1998–2000, about four months in 2007.

The common threads are narrow leadership in one story, rising rates, an energy or inflation shock, and a belief that this time is different. Several of those threads are present today.

The honest counterpoint

Breadth divergence is a condition, not a timing signal. In 1998–2000 the warning ran almost two years, and the index rose substantially in between.

In 2023 the "Magnificent Seven" carried the market on poor breadth, and participation later broadened instead of collapsing. Narrow markets can resolve with the soldiers catching up to the generals. Historically, though, the generals falling back to the soldiers has been the more common ending.

The bull case: an AI capex and earnings boom

The strongest argument against the 1999 analogy is earnings. S&P 500 profits are growing faster than at any point since 2021, powered by the AI buildout.

Measure	Latest	Source
Q2 2026 S&P 500 EPS growth	50.4% blended; 32.0% excluding Alphabet and Amazon	FactSet, Aug 7
Q2 2026 revenue growth	13.2%	FactSet, Jul 24
2026 AI investment	~$1 trillion global; just under $600 billion US	Goldman Sachs
2026 hyperscaler capex	~$646 billion, about 2% of US GDP	Apollo
Real GDP growth	1.5% in Q2; 2.1% in Q1; 0.5% in Q4 2025	BEA

Even excluding Alphabet and Amazon, this is the seventh straight quarter of double-digit growth. The headline is flattered by one-time items: Alphabet's Q2 GAAP EPS included a $98 billion gain (FactSet). Today's leaders are also highly profitable, cash-generating businesses, unlike many of the debt-financed telecom carriers of the late 1990s.

But 1999 had catalysts too. Real GDP grew above 4% a year from 1997 to 1999, corporate profits grew at double-digit rates, and a telecom and internet capex boom was sold as a productivity revolution. Cisco, the most profitable supplier of that buildout, still fell roughly 86% from its 2000 peak. Strong fundamentals extended the 1998–2000 divergence; they did not prevent its resolution.

Where we are in the cycle. Several late-cycle markers are present:

Capex is outrunning cash flow. Hyperscaler capex plus buybacks and dividends now exceeds projected cash flow, pushing these companies into debt markets (IEEE ComSoc).
End-user revenue is small relative to spending. OpenAI's $20 billion revenue run rate is roughly 3% of projected 2026 hyperscaler capex (Futurum).
The economy outside AI is soft. Real GDP grew 1.5% in Q2 while the PCE price index rose 5.1% (BEA). That two-speed economy is exactly what the breadth data shows.

One company's capex is another company's revenue. The earnings boom and the crowding are the same trade: when capex growth slows, supplier earnings reverse fastest. Goldman Sachs calls the path of AI capex growth, and when it slows, a key source of uncertainty for markets (Goldman Sachs).

What this means for advisors

The index is not the portfolio. A cap-weighted S&P 500 fund now carries heavy concentration in one theme. Many clients who believe they are diversified are not.
The experience gap is real. Clients in dividend payers, staples or small caps may be frustrated while headlines celebrate records. In 1999 that frustration peaked right before leadership changed.
Process beats prediction. Nobody can time the top. History rewards a plan for limiting losses once the trend turns, set in advance rather than mid-decline.