The assumption. Index investing rests on one sentence nobody writes down anymore: “the index is diversified.” Five hundred companies, every sector, self-rebalancing — what could be concentrated about that?

The number. As of this summer, J.P. Morgan Global Research puts the top 20 US stocks at roughly 50.8% of total market capitalization. There is no modern precedent — you have to go back half a century to find the index leaning this hard on this few names. The top 10 alone are about a third of the S&P 500. Buying “the market” now means buying the AI trade, regardless of what the remaining 480 companies do.

How cap-weighting got here. This isn’t a scandal; it’s arithmetic. A capitalization-weighted index mechanically concentrates into whatever has already won — every dollar of mega-cap outperformance increases the weight that dollar commands tomorrow. In a normal cycle the effect is mild. In a cycle where a single capital-spending theme drives the leaders — AI infrastructure capex running near $400 billion a year — the index quietly converges on one trade with one set of risk factors: chip supply, data-center buildout, the capex-to-revenue gap, and the policy rate that discounts it all.

The market knows, and doesn’t act. Bank of America’s July fund- manager survey found 45% naming an AI bubble as the biggest tail risk — up from 28% a month earlier — while flows into the same index exposure continued. July was the live demo: the Nasdaq fell almost 10% from its June peak, momentum strategies took historic damage, and by early August the index was back at an all-time high. If your “diversified” portfolio drew down and recovered in lockstep with one theme, that is data about what you actually own.

What this is not. It is not a prediction that concentration breaks. Concentrated regimes can persist for years, and the concentrated names are, on the evidence, extraordinary businesses — this earnings season 85% of S&P companies beat estimates, led by the same mega-caps. Equal-weighting or de-indexing has its own documented cost: you underperform for as long as the leaders keep leading. The point is narrower and harder: know which bet you hold, and know what would tell you it changed.

Writing the assumption down. The platform’s discipline applied to an index position looks like this — illustrations, not signals:

  • “The top-20 share of US market cap stays above 45% through year-end” — falsified by the published series.
  • “My index position behaves as an AI bet: the 60-day correlation between the cap-weighted index and its equal-weighted version stays below its 10-year average” — falsified by the paired series; if they reconverge, the concentration story is fading.
  • “A single mega-cap earnings miss moves the whole index by more than 1.5% on the day” — cascade risk, falsified (or confirmed) by the next miss.

Declare the version you actually believe, watch the series that would falsify it, and let the record — not the marketing word “diversified” — say what your index fund is.

The takeaway. Diversification is not a product feature; it’s an empirical claim about correlations and weights, and claims can be tested. The index didn’t stop being useful. It stopped being what its name implies — and the honest response is to state what it became, with a falsifier attached.

Sources: J.P. Morgan Global Research and Bank of America July 2026 Global Fund Manager Survey (both as reported by Yahoo Finance, August 2026); Reuters/FactSet earnings-season statistics, August 2026. Educational analysis — nothing here is a recommendation to buy or sell any security.

This article describes reasoning and mechanics. Nothing here is a recommendation to buy or sell any security. See the Investment Disclaimer.