The gravity of giants. What a century of data says about a market this concentrated.

1. How concentrated, exactly?

As of mid-2026, the ten largest constituents of the S&P 500 weigh close to 40 % of the index ; the Magnificent Seven alone: Apple, Microsoft, Nvidia, Alphabet, Amazon, Meta and Tesla, account for roughly 33–34 %. For perspective, the top-ten share averaged about 24 % since 1980 and stood near 18 % as recently as 2016. On a Herfindahl-type measure, the US equity market is more concentrated today than at the peak of the dot-com era in 2000, and at levels last seen in the early 1960s, when AT&T, General Motors and IBM dominated the tape (Mauboussin & Callahan, 2024).

But a weight is not a verdict. The question that matters is not whether the market is concentrated: it visibly is, but whether that concentration is anomalous relative to the economics underneath it. Here the data cuts against the reflexive bear case : Mauboussin & Callahan show that between 2014 and 2023 the top ten US stocks averaged 19 % of market capitalisation but 47 % of aggregate US corporate earnings, and by 2023, 27 % of market cap against 69 % of profits. Whatever today's giants are, they are not the profitless story-stocks of 1999.

2. The Bessembinder result: concentration is the norm, not the anomaly

The single most important paper for thinking about today's market is Bessembinder (2018), published in the Journal of Financial Economics. Examining every US common stock from 1926 onward, he found that the majority of individual stocks underperformed one-month Treasury bills over their lifetimes, and that the entire net wealth creation of the US stock market above the T-bill rate was attributable to roughly 4 % of listed companies. In the most recent update of the data (1926–2024, covering over 28,000 firms), just 2 % of companies produced about 90 % of the $79 trillion in aggregate net wealth creation.

The global replication (Bessembinder et al., 2023) is even starker : across 64,000 stocks in 42 countries (1990–2020), a little over 2 % of firms accounted for all net global wealth creation; outside the US, the majority of stocks failed to beat US Treasury bills. The distribution of long-run equity returns is not a bell curve. It is a power law with a violently long right tail: a few extreme winners pay for everything else.

This reframes the concentration debate. If nearly all long-run equity wealth has always come from a tiny handful of firms, then a capitalisation-weighted index ending up dominated by its biggest winners is not a distortion of equity investing: it is the expected terminal state of it. The index is doing precisely what it was designed to do : ride the tail.

3. Granularity: when a few firms become the macro

None of this means concentration is costless. Gabaix (2011), in Econometrica, formalised what he called the granular hypothesis : when the firm-size distribution is fat-tailed enough, idiosyncratic shocks to individual giants stop washing out in the aggregate and become macroeconomic events. He estimated that shocks to the 100 largest US firms alone explain about one third of the variation in US output growth.

Translated to today : an earnings miss at a single chipmaker is no longer stock-specific news; it is an index event, a factor event, and, through capex chains and 401(k) balances: arguably a macroeconomic one. Campbell, Lettau, Malkiel & Xu (2001) showed that the number of stocks needed to diversify away idiosyncratic risk had risen over the decades ; a top-heavy index quietly reverses that logic, because holding « 500 stocks » increasingly means holding ten risks in a trench coat. The correlation of the Magnificent Seven with each other, driven by a shared AI-capex narrative: compounds the effect: what looks like a diversified index behaves, at the margin, like a leveraged sector bet.

4. Is it a bubble? What the research actually says

The word « bubble » gets used as if it were self-evident. The literature is more demanding. Greenwood, Shleifer & You (2019) examined every US industry price run-up of 100 %+ since 1928 and found that a sharp price increase alone does not predict a crash: on average, run-ups were followed by unremarkable returns. What did predict trouble was a specific cluster of attributes : accelerating issuance of new shares, a surge in the share of young firms, extreme divergence between the run-up industry and the market, and parabolic final-stage price acceleration.

Score today's market honestly against that checklist and the result is mixed rather than damning. IPO and secondary issuance in AI-adjacent names is elevated but nowhere near 1999 levels, when Ofek & Richardson (2003) documented internet firms trading at valuations implying implausible decades of hypergrowth, sustained largely by short-sale constraints and lockup expirations. Today's leaders are mature firms with monopoly-grade margins funding capex from free cash flow. On the other hand, Pástor & Veronesi (2009) offer the uncomfortable historical regularity: during technological revolutions railways, electricity, the internet, the stock prices of innovators systematically overshoot, because investors price the new technology's expected productivity before its adoption risk resolves. High valuations during a genuine technological revolution can be individually rational and still end in a drawdown once the technology's payoff distribution narrows.

The honest synthesis : the AI build-out is real, the earnings are real, and neither fact immunises the price. Shiller's core lesson from a century of episodes is that fundamentals-backed narratives are precisely the ones that travel furthest past fair value, because they give sophisticated investors permission to stay.

5. The hidden bet inside your index fund

For the long-only investor, the practical issue is not forecasting the top, nobody does that reliably: but recognising what a cap-weighted allocation has silently become. Three research-backed observations :

• Buying the biggest has historically carried a toll. Arnott & Wu found that the #1 company in each sector underperformed its sector average by roughly 3 % per year over the following decade, across countries, leadership at the top of the capitalisation table has historically been mean-reverting, even while the market itself rose.
• Concentration changes your risk, not just your return. With 40 % of the index in ten names sharing one macro narrative, tracking-error thinking inverts : the passive investor is now the one making the concentrated bet, and the diversified investor is the one « deviating ».
• Naive diversification is not naive. DeMiguel, Garlappi & Uppal (2009) showed that a simple equal-weight (1/N) allocation is remarkably hard to beat out-of-sample once estimation error is accounted for, a rare piece of quantitative support for the humblest possible response to concentration risk.

6. How a serious investor responds (without calling tops)

1. 01Separate the business bet from the index bet. Decide explicitly how much exposure you want to the AI-capex complex, then check what your « neutral » cap-weighted holdings already give you. For most index investors the honest answer is : far more than they chose. Blending cap-weight with equal-weight or capped-weight exposure restores the decision to you without requiring a market call.
2. 02Watch the Greenwood–Shleifer attributes, not the price. Run-ups don't predict crashes ; issuance surges, new-firm floods and parabolic acceleration do. Those are observable, updatable indicators, a checklist, not a feeling. As long as the giants' earnings share exceeds their index weight, the situation is structurally unlike 2000 ; if that inverts, re-underwrite.
3. 03Respect the tail: in both directions. Bessembinder's data cuts against both the permabear and the concentration-chaser : missing the handful of extreme winners is catastrophic for long-run returns, and so is holding yesterday's #1 on autopilot. The defensible posture is breadth with deliberate, sized tilts, never a binary in-or-out on the market's largest names.

7. The bottom line

Today's concentration is historically extreme, but the research does not support reading it as a simple bubble signal. Equity wealth has always been created by a violently small number of firms ; cap-weighted indices are built to capture exactly that skew ; and today's giants earn a share of profits even larger than their share of the index. What concentration does change is the risk architecture of every portfolio benchmarked to the index : more granular, more narrative-correlated, more exposed to a single technology's adoption curve than at any point since the dot-com era, with the historical record showing that the biggest names, specifically, tend to lag from the top.

The investors who treat concentration as a market-timing signal will likely be wrong twice: once on the way up, once on the way back in. The ones who treat it as a portfolio-construction problem, measuring the bet they actually hold, diversifying it deliberately, and monitoring the specific attributes that have historically separated revolutions from manias are the ones the data quietly favours.