Market valuation briefing · P/E and P/S, by category and over time
The S&P 500 trades at 19.4x forward earnings, right at its 10-year average, so on what analysts expect, the market looks ordinary. On what companies have actually delivered, it does not: trailing P/E is 26.15x against a long-run mean of 16.23, price-to-sales is 3.78x, an all-time high, and the Shiller CAPE at 41.07 is the second-highest reading ever recorded. This page breaks those multiples down by sector, by size and style, and across two decades of history.
Vintages: sector P/E Sep 30, 2026 · sector forward sales Aug 24, 2026 · style indexes Aug 31, 2026 · index and history data Oct 1, 2026 · within-technology cuts Jul 2026 to Oct 1, 2026 (see Section 02 notes) · Compiled Oct 2, 2026
Black tick on each track marks the long-run or 10-year average. Teal means near or below that average; rust means far above it.
On expectations, yes: the S&P 500 trades at 19.4x forward earnings, right at its 10-year average. On results, no: trailing P/E is 26.15x against a long-run mean of 16.23, price-to-sales is 3.78x, an all-time high, and the Shiller CAPE at 41.07 is the second-highest reading ever recorded. That forward line is fair only if analysts' earnings forecasts land, so the valuation debate is really about whether forward earnings are real.
Technology at 21.1x forward is no longer the market's expensive sector; Industrials at 22.6x and Consumer Staples at 22.0x now outrank it. Energy at 13.1x, Financials at 14.1x and Utilities at 15.5x are the only sectors priced meaningfully below the index. Within technology, software at 15.8 to 15.9x trades below the index while semiconductors carry the premium, and the AI premium sits in story names rather than the largest earners.
Three things: trailing P/E and P/S drifting back toward historical norms because reported earnings rise with prices roughly flat; AI delivery spreading past semis and mega-caps into enterprise software and IT services; and the direction of revisions. If the forward P/E rises because estimates are being cut while prices hold, the bull case is getting more expensive, not cheaper.
All 11 S&P 500 sectors on forward earnings, then on sales. Toggle between the two lenses: the ranking reshuffles completely, because earnings multiples punish low-margin businesses while sales multiples punish low-revenue ones. Real Estate looks cheap on earnings and dear on sales; Technology is the mirror image.
Technology at 21.1x forward is no longer the market's expensive sector. It trades below its own recent history (28.2x in Dec 2021, 28.4x through 2024-25), while Industrials (22.6x) and Consumer Staples (22.0x) now outrank it. Defensive staples costing more than tech is not the usual order of things.
Energy (13.1x), Financials (14.1x) and Utilities (15.5x) are the only sectors priced meaningfully below the index. On sales, the cheapest dollars of revenue sit in Staples (1.5x), Energy (1.6x) and Health Care (1.7x); a dollar of Technology revenue costs 7.2x, and Real Estate revenue 6.4x.
Technology at 21.1x forward is an average of businesses priced nothing alike. Combined software (Zacks Computer-Software) at 15.8 to 15.9x trades below the S&P 500 itself, and the narrower IT-services cut is lower still at 13.6x, the lowest industry multiple cited here. Semiconductors, where the AI earnings actually are, carry the premium. Toggle between forward earnings and sales below; every bar carries its methodology, because in this corner of the market the method moves the number more than the market does.
Software at 15.8 to 15.9x forward (Zacks, Jul 2026) sits below the index, and far below earlier readings on a different (cap-weighted) Yardeni basis: Application Software at 27.8x in Jun 2023 and 49.2x in Aug 2021, shown for scale rather than as a like-for-like discount. Treat the July figure as stale-high; the selloff continued into September. The market is pricing AI as a headwind for application software, the opposite trade from semis.
Semiconductors show 37.0x on a Zacks company-average (Aug 2026) and 23.4x cap-weighted (Yardeni, Feb 2026). Both are on the chart, labeled, because neither alone is honest: the average is lifted by small story names, the cap-weighted figure is dominated by the giants. Against a 5-year average of 15.4x (Morningstar/PitchBook, cap-weighted), semis are expensive on either method.
Forward P/E ranges, next-12-month consensus (FinanceCharts, Oct 1, 2026, cross-checked against GuruFocus, MarketBeat and others; sources disagree where fiscal-year conventions differ, so ranges are shown, not points). The spread inside this one list, from Micron at about 6x to Palantir at 79 to 109x, dwarfs the entire sector-versus-market spread in Section 01. Tesla (175 to 422x) is off the scale and annotated, not charted.
Sorted cheapest to most expensive by range midpoint. Nvidia at 24.9 to 25.6x trades below Apple at 32.1 to 35.9x and near its own 25.87x median: the company earning the most AI money is not the company priced for the most AI hope. AMD at 32 to 69x runs at two to three times Nvidia's multiple despite Nvidia being the bigger AI earner (methodologies differ widely across sources for AMD, hence the broad range).
IT services is tech's value shelf: 18.48x on the broad Zacks cut and 13.62x on the narrower services cut (Sep 2026), about 19 to 33% below its own 5-year medians of 27.63x and 16.89x respectively, at about 1.24x trailing sales (NYU Stern, aggregate basis, Jan 2026). If AI turns out to be an implementation boom rather than a replacement, this is the cheapest way the market lets you bet on it.
Hardware at about 27.5x and electronic equipment at 26.2 to 28.3x look dear against the index, yet sit right at their own 5-year medians (25.96x and 29.36x respectively, Zacks company-average basis). Expensive versus the market is not the same as expensive versus their own past; both statements are true here, and the second one matters.
Alphabet (16.6 to 16.9x) trades below both the S&P 500's 19.4x and the tech sector's 21.1x; Broadcom (18.1 to 21.0x) straddles the index line while staying below the sector; Meta (19.7 to 22.6x) straddles the sector line, and Nvidia (24.9 to 25.6x) sits above both benchmarks yet below Apple (32.1 to 35.9x) and far below story names like Palantir. The premium sits in semis broadly, semi-equipment names like Lam Research (33.8 to 36.0x), and story stocks, not concentrated in the mega-cap earners.
Micron at 6.2 to 6.6x, Adobe at 8.7 to 10.3x and Intuit at 10.8 to 12.2x sit at one end; Palantir at 79 to 109x at the other. That roughly fifteen-fold spread inside technology is several times the spread between the cheapest and dearest sectors in Section 01. Owning "tech" says far less than it used to about what you paid.
At 15.8 to 15.9x, with Salesforce at 14.4 to 20.1x and ServiceNow at 27.0 to 34.2x, both well below their own three-to-five-year multiples, the market has already voted that AI deflates seats. If AI instead proves an enterprise ARPU driver, this is the opportunity; if the disruption is real, it is the value trap. The data here cannot settle that, and neither can one quarter of earnings.
Open gaps, stated plainly: there is no clean standalone forward P/E for semiconductor equipment or for Interactive Media & Services as an industry (the sector anchor is Siblis Communication Services at 17.80x forward, Aug 2026; company-level P/S runs 8.0 to 8.6x for Meta and 9.2x for Alphabet, late Sep 2026). Price-to-sales methodologies diverge four- to six-fold on the same industry (NYU Stern aggregate versus GuruFocus medians), so the sales toggle above uses one method only and labels it.
Forward multiples fall steadily as you move down the size ladder, and the growth premium over value is wide but narrower than its own history. Price-to-sales is not published for the mid, small and style indexes, so that lens covers the S&P 500 only.
Forward estimates: S&P 500 via FactSet, Oct 1 2026; MidCap 400 and SmallCap 600 via Yardeni Research, early Sep 2026; Russell 1000 styles via Siblis Research, Aug 31 2026. Trailing P/E for the 400 and 600 via ETFdb (IJH, IJR). Methodologies differ slightly between providers, within roughly 0.3x on cross-checks.
The S&P SmallCap 600 at 15.1x forward trades about 22% below the S&P 500's 19.4x, and the MidCap 400 sits between at 15.8x. On trailing earnings the order flips around: small caps at 19.81x are the cheapest of the three size indexes on results already delivered.
Russell 1000 Growth runs 23.27x forward and 35.6x trailing, against Value at 16.83x and 22.48x. The trailing growth-to-value ratio of 1.58x is below its 10-year median of 1.71x, and Value's own CAPE of 25.73 is the highest in its series: value is cheap relative to growth, not cheap in absolute terms. Growth's CAPE is 53.43.
Switch between the three lenses. Trailing P/E whipsaws whenever earnings collapse (2008, 2020), which is why the CAPE, built on ten years of inflation-adjusted earnings, exists. Price-to-sales ignores earnings entirely, and its story is the starkest: a near-unbroken climb from 0.87x in 2008 to an all-time high of 3.78x now.
Forward P/E anchor points from FactSet Earnings Insight editions across the decade. Coverage is uneven because the reports typically named only the highest and lowest sectors each period, so treat this as signposts, not a continuous series. Toggle sectors on and off; the dashed line is the S&P 500 itself.
All points are estimated snapshots transcribed from report text, each with its own as-of date between 2016 and Sep 2026; 2019 is a gap year with no published anchors. Energy's 2016-17 readings (31.5x, 28.3x) are an artifact of the 2015-16 oil crash collapsing its earnings denominator, not evidence the sector was ever priced like growth. Communication Services appeared as Telecom Services before the 2018 GICS reclassification. Real Estate has no published anchors. The sector values here can differ slightly from the Sep 30, 2026 figures in Section 01 because the snapshot dates differ.
Forward P/E at its 10-year average (19.4x vs 19.0x) means today's price is fair only if analysts' earnings forecasts land. Trailing P/E at 26.15x and CAPE at 41.07, the second-highest ever, say there is little cushion if those forecasts slip. The valuation debate is really a debate about whether forward earnings are real.
Technology at 21.1x forward is below its own 2021 and 2024-25 levels, while Industrials (22.6x) and Staples (22.0x) top the table. The premium has rotated toward industrials and defensives. On sales, though, tech remains in a league of its own at 7.2x forward revenue, with Real Estate next at 6.4x.
Small caps trade about 22% below large caps on forward earnings, and value's relative discount to growth is narrower than its 10-year norm, with value's own CAPE at a series high. Neither small caps nor value are cheap against their own history in absolute terms; they are cheap only against a large-cap index at record sales multiples.
Sections 01 to 06 say what the market costs. How alarming that is turns on one question: is the earnings base about to change faster than history suggests? The bull case says that is exactly what AI is doing, and that the extreme trailing multiples measure price against earnings that are about to be revised upward. The bear case says the forward line is a promise built on optimistic estimates, and that the delivery so far is concentrated in a handful of names, while anyone buying the index pays the cap-weighted aggregate multiple across all 500. The sharper version of both: the market is not pricing euphoria, it is pricing a broadening, and each side below is written entirely from the numbers already on this page.
The clean resolution: trailing P/E and P/S drift back toward their historical norms because reported earnings and revenue rise, with prices roughly flat. That is the market growing into its multiple, and it is observable quarter by quarter in reported results rather than in forecasts.
The index multiple becomes self-supporting if AI delivery spreads past semis and the mega-caps: enterprise software monetizing AI as higher per-seat pricing rather than losing seats, IT services winning implementation demand. Section 02 is where that test is already visible in prices.
If the forward P/E rises because estimates are being cut while prices hold, the bull case is getting more expensive, not cheaper. If it stays near 19x while trailing EPS climbs toward it, the earnings are doing the work. The direction of revisions matters more than any single reading.