The Magnificent Seven is often discussed as a single factor bet — a concentrated slice of mega-cap growth that has driven index returns and crowded factor portfolios alike. Over the past three years, that story held: NVDA alone returned more than 318%, the group powered cap-weighted indices, and anything that wasn’t mega-cap tech felt like a lagging trade. Strip away the label, though, and what you find is not one homogeneous bet but seven very different ones — with different return drivers, different appropriate valuation lenses, and, as 2026 is now showing, very different near-term momentum.
Return Dispersion: One Winner, One Laggard, and a Wide Middle
Click chart to expand
The top panel of the chart makes the dispersion impossible to ignore. NVDA has returned more than 318% since September 2023 — roughly triple the next-best name, GOOG at 153%. The MAGS ETF, which packages the group into a single ticker, sits at 108% — a respectable outcome, but one that masks the underlying spread.
Below NVDA and GOOG, four names cluster in a tighter band: AMZN (76%), AAPL (72%), META (71%), and MSFT (37%). At the bottom sits TSLA at just 15% — a name still counted among the Mag 7 in most factor frameworks, but one that has delivered a fraction of what its peers have over this period.
For allocators treating “Mag 7 exposure” as a homogeneous growth tilt, this dispersion matters. Equal-weighting the group would have looked very different from cap-weighting it — and either approach would have told a different story than simply owning the S&P 500’s top seven names.
The Right Multiple for the Right Business
Comparing Mag 7 valuations side by side only works if you use the right yardstick for each business model — and the chart’s middle and bottom panels are most informative when read that way. Forward P/E is the natural metric for high-margin, asset-light platforms where earnings scale smoothly with revenue: software, advertising, and cloud services. Forward P/S is often the better lens for cyclical, capital-intensive, or low-margin businesses where earnings swing wildly through the cycle and a single year’s profits can flatter or punish the multiple.
That distinction explains much of what looks like “valuation inversion” in the data.
Where P/E Works — and Where It Misleads
The middle panel is most reliable for the platform and software names. META at 17.6x forward earnings, MSFT at 23.0x, and GOOG at 24.9x are all businesses where margins are structurally high and earnings are a reasonable proxy for economic value creation. AMZN at 25.9x sits in this band too, though its blended retail-and-cloud model makes P/E alone an incomplete read. MAGS aggregates to 23.9x — a reasonable earnings multiple on the surface, but one that blends businesses for which P/E means very different things.
NVDA is the case study in why context matters. At 19.5x forward P/E, it screens as the second-cheapest name in the group despite delivering the best returns — because earnings have exploded alongside the AI buildout and the denominator has caught up to the numerator. That is a real signal: the market is not pricing NVDA on hope alone at this point. But chip businesses are cyclical by nature. Earnings can compress as fast as they expand when capacity catches up, mix shifts, or pricing normalizes. For semiconductors, P/E at a cycle peak can look deceptively cheap — which is precisely why many investors anchor on P/S through the cycle instead.
At the other extreme, TSLA’s 152x forward P/E is almost unusable as a comparative metric. Earnings are too volatile, too policy- and volume-dependent, and too far from what the market is actually underwriting (autonomy, energy, robotics) for a single-year earnings multiple to tell a clean story. AAPL at 35.7x sits between the two worlds — a hardware business with a growing services annuity, where P/E is informative but never the full picture.
Why P/S Is the Better Lens for Cyclicals and Low-Margin Models
The bottom panel is where the business-model differences become clearest. On forward price-to-sales, NVDA (12.0x) and TSLA (11.5x) lead the group — a very different ranking than P/E implies. For NVDA, the elevated P/S reflects what the market is actually paying for: revenue growth and share of the AI compute stack, independent of where margins sit in any given quarter. The sharp step-down in NVDA’s P/S line around early 2025 is telling — a projected surge in revenue compressed the multiple even as the stock kept climbing. That is classic cyclical math: when the top line scales, the sales multiple falls even if the market’s conviction hasn’t.
TSLA’s 11.5x P/S is more comparable across the group than its 152x P/E — but it still places the stock near the top of the group on a revenue basis, despite delivering the weakest price return over this period. The market is paying a premium for dollars of sales, not dollars of earnings.
At the other end, AMZN at 3.1x sales looks like a completely different animal — and it is. Retail compresses margins structurally, so revenue multiples stay low even for a dominant franchise. GOOG (3.7x) and META (4.7x) also screen as inexpensive on P/S, which is the flip side of their high-margin economics: every dollar of sales converts efficiently to profit, so the market doesn’t need to pay up on revenue the way it does for NVDA’s cyclical top line. AAPL (10.2x) and MSFT (8.6x) sit in the middle, consistent with hardware-software and enterprise software models respectively. MAGS blends to 7.8x — a figure that averages businesses whose appropriate valuation lens differs fundamentally.
Read together, the two valuation panels argue against ranking the Mag 7 on a single multiple. META is cheapest on P/E; AMZN is cheapest on P/S; NVDA looks cheap on earnings but expensive on sales — and for a chip company at this point in the cycle, both readings are telling you something different. That is not inconsistency in the data. It is the point.
YTD: Breadth Is Telling a Different Story
The long-run dispersion and valuation nuance above help explain why the Mag 7 trade was never as clean as the headlines suggested. The year-to-date picture helps explain what is happening now — and it is a meaningful shift.
Click chart to expand
Through late July, RSP — the Invesco S&P 500 Equal Weight ETF — has returned +12.6%, outpacing SPY at +9.1%. MAGS, the dedicated Mag 7 basket, is -2.2%. After years in which cap-weighted mega-cap growth pulled further and further ahead of the broader market — and equal weight consistently felt like the trade that couldn’t catch a bid — 2026 is running the other way.
The breadth panel makes the mechanism visible. Roughly 63% of SPY’s holdings are trading above their 50-day moving average — healthy, broad participation across the index. Within MAGS, only 37.5% of the group’s constituents are above their 50-day trend. Mag 7 breadth spent much of the first quarter near zero, spiked briefly when all seven names aligned in late May, and has rolled over sharply through July. The group is not moving as a unit — and on balance, more of it is weakening than strengthening.
That is consistent with the longer-term dispersion story. When NVDA carries the group but TSLA, MSFT, and others lag, cap-weighted Mag 7 exposure and cap-weighted index exposure can diverge sharply from equal-weighted broad market results. Equal weight doesn’t care which seven names dominate the narrative. It cares whether the other 493 stocks are participating — and in 2026, they are.
The So What
The Mag 7 has been the dominant factor story of this cycle, but treating it as a single flavor oversimplifies what is actually in the bottle. NVDA’s long-run outperformance has been extraordinary, and its 19.5x forward P/E confirms earnings have scaled with the AI trade — but its 12.0x P/S reminds you the market is still paying up for cyclical revenue, not just current profits. TSLA’s 152x P/E is noise; its 11.5x P/S is the more honest comparison. META looks cheapest on earnings; AMZN looks cheapest on sales — two valid reads that point to different conclusions depending on which lens you apply.
Zoom out to 2026, and the factor implications sharpen. A Mag 7 tilt that worked brilliantly for three years is underwater year to date while equal-weighted U.S. large cap is leading. Breadth confirms it is not just a sizing issue — participation within the group is deteriorating even as the broader index holds up. For advisors who spent years explaining why equal weight lagged, the current regime is a reminder that concentration cuts both ways.
For ETF investors, products like MAGS offer convenient access but smooth over the dispersion that defines the group. The 108% cumulative return since 2023 is real, but it is an average of seven very different outcomes — and the -2.2% YTD return is what happens when that dispersion turns against you at the wrong time. The Mag 7 is not one factor. It is seven flavors — and in 2026, the rest of the market is starting to pull its weight again.
📈 Interactive Analysis: Explore Mag 7 holdings, valuations, and factor exposures on ETF Action.
Disclosures
This material is for informational purposes only and should not be considered investment advice. All investments, including ETFs, involve risk, including the possible loss of principal.
This analysis was developed by the team at ETF Action. We leverage advanced AI tools to assist in the drafting and refinement of our content, based on our expert prompts, direction, and final review.
