Market leadership / September 2026
AI stock rotation in 2026:
the broadening lost momentum
Smaller companies and equal-weight stocks had a real lead. September has tested it. Here is what the returns, earnings and financing conditions say about the next phase of the AI trade.
By Jaber Khalifa · ApexTicker
Updated · 11-minute read
The verdict
Broadening earlier in the year. Narrower leadership in September.
Yes, the market broadened beyond mega-cap technology in 2026. No, the latest evidence does not support a lasting exodus from AI. Equal weight’s earlier advantage has reversed, small-cap leadership has weakened, and AI suppliers still show substantial business growth. The investment question is which earnings and cash flows justify their price.
Evidence cutoff: S&P total returns through September 24; major-index price returns through September 25. These are separate, explicitly dated comparisons. Sources & calculation notes ↓
01 / Follow the relative returns
The same market tells two different stories
“AI rotation” can mean two things: more stocks sharing in a rally, or AI-related leaders being displaced. Those are different claims. The cleanest first test is to compare the S&P 500 with its equal-weight counterpart: the same companies, but a different distribution of influence.
Through August 31, equal weight returned 15.59% against 13.14% for the conventional S&P 500. The Russell 2000 returned 20.01%. Those gains support an earlier broadening. Clark Capital’s August benchmark review, using Morningstar data, reports these index returns.
August 31 · equal weight ahead
+2.45 ppSeptember 24 · equal weight behind
−2.29 ppApexTicker calculations: equal-weight minus cap-weight YTD total return, in percentage points (pp). August: Clark Capital / Morningstar. September: S&P DJI, September 24 dashboard, p. 1.
By September 24, that relationship had reversed:
| Index | 2026 YTD | September* | Q3 to date |
|---|---|---|---|
| S&P 500 | +13.51% | +0.33% | +3.00% |
| S&P 500 Equal Weight | +11.22% | −3.78% | −0.82% |
| S&P SmallCap 600 | +15.59% | −4.33% | −6.71% |
*August 31 close to September 24 close; not a completed month. Dividends reinvested, before fund fees. Source: S&P Dow Jones Indices, p. 1. SmallCap 600 and Russell 2000 are different benchmarks.
Change the window. Change the story.
Same three indexes · same September 24 cutoff
The lead accumulated earlier in a year can survive after leadership has turned. A positive small-cap YTD number is therefore insufficient evidence that the rotation is strengthening now.
Friday’s close reinforces the distinction
At the September 25 close, the Russell 2000 was up 14.3% for the year, the S&P 500 13.1%, and the Nasdaq Composite 16.5%. That week, small caps lost 0.8% while the S&P gained 1.2% and Nasdaq gained 2.1%. These are price returns from the Associated Press closing report, republished by AM 590; do not subtract them from the total-return table above.
Small caps still lead the S&P on this YTD measure, but trail Nasdaq. Nasdaq Composite is a technology-heavy market benchmark, not a pure AI portfolio. The evidence supports a qualified broadening story, not a clean handoff from AI to everything else.
02 / Look beneath the index
Energy leadership is not the same as broad participation
Through September 24, only three S&P sectors were positive for the month. Energy led YTD, while technology also beat the index. S&P’s sector dashboard shows the split:
All 11 S&P 500 sectors · September 24 snapshot
| Sector | 2026 YTD | September |
|---|---|---|
| Energy | +42.50% | −1.18% |
| Information Technology | +27.61% | +3.85% |
| Health Care | +11.16% | +0.02% |
| Materials | +10.80% | −5.04% |
| Industrials | +9.81% | −3.30% |
| Consumer Staples | +7.35% | −2.01% |
| Real Estate | +7.28% | −4.38% |
| Communication Services | +6.57% | +6.41% |
| Financials | +0.85% | −5.13% |
| Consumer Discretionary | −4.53% | −4.53% |
| Utilities | −5.80% | −6.05% |
Sorted by YTD return. Source: S&P DJI, September 24, 2026, p. 3. Sector participation is not the percentage of individual stocks advancing.
The distinction matters for portfolio construction. An energy holding can respond to oil supply and pricing while a chip supplier responds to computing demand. Both can outperform for different reasons. That does not mean the entire nontechnology market has become stronger.
AI exposure also crosses sector boundaries. A data-center equipment supplier can share a customer's spending cycle with a semiconductor business even when their sector labels differ. Buying more sectors does not automatically diversify the economic driver. Our guide to correlation and diversification explains how to examine that overlap.
What equal weight actually measures
The equal-weight S&P 500 holds the same constituents as the cap-weighted index, resetting company weights equally each quarter. It reduces the influence of the largest businesses, but introduces different size and sector exposures. It is a useful participation proxy—not an AI-free index or a count of advancing stocks. S&P index definition.
03 / Separate profits from prices
Earnings have broadened. The quality of those earnings matters.
The case for looking beyond the biggest technology companies has a fundamental basis. FactSet’s August 28 report put Q2 blended earnings growth for the 493 S&P companies outside the Magnificent Seven at 31.8%. That is evidence of profit growth beyond the headline group, although it is an aggregate—not proof that every constituent is improving.
The same report placed Magnificent Seven earnings growth at 118.5%, but highlighted large investment gains at Alphabet and Amazon. Excluding those two companies brought growth for the remaining five to 43.2%. That exclusion changes the group; it is not a normalized growth rate for all seven. FactSet also uses a mixture of GAAP and non-GAAP EPS according to analyst reporting conventions. Read FactSet’s earnings breakdown.
The useful conclusion: earnings growth outside mega-cap leaders can coexist with continued AI strength. Neither the size of a headline earnings increase nor a low multiple based on it tells you how much recurring cash the business produces.
Forward estimates are improving unevenly
FactSet’s September 4 analysis found that the Q3 bottom-up EPS estimate rose 1.2% between June 30 and August 31, yet only four of 11 sectors saw increases. Seven saw cuts. The aggregate improved without a majority of sectors improving. These are analyst estimates, not reported Q3 earnings. Source: FactSet’s estimate-revision analysis.
AI demand has not disappeared
NVIDIA
Fiscal Q2 2027 · ended July 26, 2026
$96.2bnQuarterly revenue, +106% year over year. Data Center revenue: $89.0bn, +117%.
Microsoft
Fiscal Q4 2026 · ended June 30, 2026
+43%Year-over-year Azure and other cloud services revenue growth. Company operating income: $40.6bn, +18%.
These reports show demand at a supplier and a cloud platform. Azure includes non-AI services, and neither release establishes the return earned by every customer buying AI capacity. Revenue at one link in the chain can grow before another link proves its economics.
The valuation question: what must go right?
A rotation thesis needs a price discipline. Before calling an AI leader expensive or a smaller company cheap, compare the same earnings period, separate operating profit from investment gains, and account for debt and capital spending. Otherwise, the apparent discount may be an accounting difference or compensation for a weaker balance sheet.
If recurring EPS grows 20% while the valuation multiple falls 20%, the implied share-price change is 1.20 × 0.80 − 1 = −4%, before dividends. Excellent business growth can still produce a disappointing stock return. The same arithmetic works against buying a low-multiple company whose earnings are about to shrink.
For company research, pair the earnings-report guide with free cash flow versus net income. The goal is to identify what remains after one-off gains and the investment needed to sustain growth.
04 / The financing test
The rate-cut shortcut no longer fits September
On September 16, the Federal Reserve raised its target range by 0.25 percentage point to 3.75%–4.00%. Its statement described inflation as elevated. A small-cap thesis built simply on imminent policy easing therefore needs to be reconsidered against the actual decision. Source: Federal Reserve statement.
That does not establish the cause of every September stock move. It does change the test for businesses that depend on external finance:
- Debt timing: how much borrowing matures before the business can generate cash internally?
- Rate exposure: is debt fixed or floating, and what would refinancing cost at current terms?
- Funding flexibility: can management slow capital spending without damaging the business, or would it need to issue equity?
Company size alone cannot answer these questions. A profitable small business with net cash and a debt-heavy AI infrastructure operator can have the opposite risk profile from the usual “small caps versus big tech” stereotype. Our guide to rates and stock valuations covers the mechanisms.
05 / Turn the evidence into decisions
Diversify the risk you own—not the headline you fear
The evidence does not justify an automatic “sell AI, buy small caps” switch. It does justify checking whether your portfolio needs every part of the AI spending cycle to go right at once.
- Look through the fund names.
Add direct holdings to the same stocks held inside broad-market, technology and thematic ETFs. Then identify shared customers and spending drivers. Our ETF diversification guide explains why adding funds can leave concentration largely intact.
- Separate rebalancing from a new forecast.
Trimming a position that exceeds your chosen allocation is a risk-control decision. Increasing small caps because you expect them to outperform is an active bet. Write down which decision you are making and the evidence it requires.
- Underwrite the replacement.
A business does not become attractive because it is outside AI. Check recurring earnings, cash conversion, leverage and the purchase price. For small caps, inspect whether an index or fund includes loss-making firms and how its valuation statistics treat them.
- Choose an implementation you can sustain.
Review transaction costs, taxes and liquidity before changing holdings. New contributions may help restore target weights without an immediate sale. Position size should reflect your time horizon and ability to absorb losses, not confidence in a market narrative.
The strongest reason to reduce an AI position is that its price, business risk or portfolio weight no longer fits your plan. A short period of outperformance elsewhere is a weaker reason. A diversified allocation can accommodate AI growth and opportunities beyond it.
06 / A view that can be tested
What would turn broadening into a durable rotation?
Use the same measures at each review. Relative returns describe leadership; earnings and financing help judge whether it can last. The following are conditional scenarios, not predictions or trading signals.
| Scenario | What to look for | What it would mean |
|---|---|---|
| Broadening resumes | Equal weight and small caps regain relative strength across several months; more sectors receive earnings upgrades. | A stronger case that participation is expanding with fundamental support. |
| AI leadership persists | Cap weight keeps leading; AI revenue converts into sustainable cash flow; other sectors’ revisions remain mixed. | Earlier diversification gains have not displaced the existing leaders. |
| Market-wide pressure | Both groups fall; earnings estimates weaken and refinancing becomes harder. | Losing less is relative strength, not evidence of a healthy rotation. |
Keep one-, three- and six-month relative-return windows, matching dates and dividend treatment. Add sector-level earnings revisions and debt conditions. A rebound without those supporting changes deserves less confidence than a trend confirmed across them. Our sector rotation guide develops the broader framework.
Questions investors are asking
Is AI dominance ending in 2026?
The evidence supports earlier broadening followed by renewed concentration in September. That is a change in market leadership, not proof that the AI business cycle has ended. Strong reported AI-related demand can coexist with poor returns for an overpriced stock.
Does equal weight outperforming prove money is leaving AI?
No. It shows a different weighting of the same companies performed better. It does not identify investors’ transactions or where capital came from. A fund-flow claim requires a separately sourced flow series with a defined universe and period.
Are small caps the safer alternative?
Not automatically. Size is not a measure of balance-sheet strength, liquidity or valuation. A switch may reduce exposure to mega-cap stocks while increasing refinancing risk and sensitivity to the economy.
Why are the September 24 and 25 returns different types?
The S&P dashboard supplies a consistent set of total returns, including dividends, for September 24. The AP report supplies major-index price returns for September 25. Each comparison stays within its own date and methodology; the figures are not spliced into one series.
Sources & methodology
Scope: U.S. equity leadership, with evidence published by September 28, 2026. The main comparison uses S&P DJI’s September 24 snapshot because it provides consistent index and sector total returns. The September 25 AP close supplies a separate, later price-return check.
Periods: YTD begins at the December 31, 2025 close; Q3 begins at the June 30, 2026 close; September begins at the August 31 close.
Calculations: relative leads are arithmetic differences in percentage points: 15.59 − 13.14 = +2.45 at August-end; 11.22 − 13.51 = −2.29 on September 24. They are not relative-return ratios. The EPS/multiple example is hypothetical. No fund flows, stock-level breadth series or live valuation multiples are inferred from these returns.
Limits: AI has no single universally representative benchmark. Equal weight, sector indexes and small-cap indexes test different aspects of leadership. Positive sector counts are a limited breadth measure. Company fiscal periods differ; analyst estimates can change. Index returns exclude investment-product costs. Past performance does not establish future leadership.
- S&P Dow Jones Indices — Daily Index Dashboard (PDF)September 24, 2026 edition; p. 1 index returns, p. 3 sectors. USD total returns. The publisher’s rolling URL may subsequently show a newer edition; the observation date used here remains September 24.
- Clark Capital — August 2026 Benchmark ReviewPublished September 4, 2026; Table 1, Morningstar-sourced August-end index returns. Used for the earlier broadening comparison.
- Associated Press — September 25 market closeRepublished by AM 590 The ANSWER. Major-index YTD and weekly price returns.
- FactSet — Magnificent Seven earnings and investment gainsAugust 28, 2026. Q2 earnings, the other 493 companies, and accounting caveats.
- FactSet — Q3 EPS estimate revisionsSeptember 4, 2026. Changes from June 30 through August 31; forecasts, not actual Q3 results.
- NVIDIA — Fiscal Q2 2027 resultsReleased August 26, 2026; quarter ended July 26, 2026.
- Microsoft — Fiscal Q4 2026 resultsReleased July 29, 2026; quarter ended June 30, 2026.
- Federal Reserve — September 16 FOMC statementPolicy-rate decision and stated inflation assessment.
- S&P DJI — Equal Weight Index definitionConstituents and quarterly equal weighting.
Originally published . Substantially rewritten to replace the original “end of AI dominance” thesis with dated market evidence.
This analysis is educational, not a personalized recommendation. Investments can lose value, and diversification does not eliminate market risk. Read the full disclaimer.