Predictium · Bubble Watch
Mag 7 module · weekly
As of 17 Aug 2026
Drill-down · the index epicenter

Magnificent 7 Heat

62/100
Hot — but earnings-backed
diversified · cheapconcentrated · manic

Record concentration and a ~$700B capex arms race run hot — but the valuation premium is the smallest in a decade and the earnings are real. The risk here isn't the multiple. It's whether the AI spend earns its return.

Weight in S&P 500
~33%
near record concentration
2026 capex (AI)
~$700B
up from ~$400B in '25
Fwd P/E premium vs rest
~+45%
vs +100% in 2000
Share of S&P earnings
~30%
vs <20% at '00 peak
The 5 factors behind the 62
Concentrationhigher = hotter
90
~33% of the S&P in 7 names · most concentrated index in 50+ years
AI capex intensityhigher = hotter
80
~$700B/yr, near-parabolic · ROI still unproven (the modern telecom-overbuild risk)
Internal dispersionhigher = hotter
70
Tesla ~190× vs Microsoft ~27× · widening spread is a classic late-cycle tell
Valuation premiumlower = cooler
40
~45% premium to the rest of the market — the lowest in a decade, far below 2000's 100%+
Earnings qualitylower = cooler
30
~30% of all S&P earnings, funded by real cash flow — the opposite of 2000's profitless dot-coms
Name by name — same basket, different risk
CompanyFwd P/E*HeatRead
Tesla TSLA~190×The bubble outlier — priced on robotaxi/robotics dreams
Nvidia NVDA~24×Largest weight; fwd P/E reasonable vs its growth
Amazon AMZN~30×Capex-heaviest (~$200B); cloud is the payoff
Apple AAPL~30×Light on AI capex; AI strategy still the question
Microsoft MSFT~27×Cheapest hyperscaler; capex leash tightening
Meta META~29×Capex visibly converting to ad revenue
Alphabet GOOGL~26×'26 leader; cloud growth topping forecasts

Why this isn't 2000

At the dot-com peak the biggest names carried a 100%+ valuation premium on under 20% of market earnings. Today it's a ~45% premium on ~30% of earnings, paid for with real cash flow. The genuine tail risk has moved: it's no longer the multiple, it's ~$700B a year of AI capex chasing a return that hasn't fully shown up yet — plus Tesla-style single-name froth inside the basket.

Get the weekly read

Monday mornings — the number + what moved. Free, never gated.
*Forward P/E figures are approximate, mid-2026, blended from multiple sources and move daily.
Sources: Stock Analysis / Motley Fool (S&P weight ~33%), Yahoo Finance (2026 capex ~$680–750B), Goldman Sachs via Columbia Threadneedle (premium & earnings-share comps).
Methodology: factor scores 0–100 by position vs history; the 62 is an equal-weight blend shown illustratively. Feeds the Concentration bucket of the main Bubble Index.
Methodology & sources: how every score is built →
Not investment advice. A high reading signals fragility and weak long-run expected returns — not a near-term crash call.