Key takeaways
- The top 10 index weight in the MSCI World Index was 26.61% on 31 August 2026, against 9.77% in October 2015, while the index shed 362 constituents.
- Two of those ten lines are share classes of one company. Alphabet A at 2.15% and Alphabet C at 1.69% make the list nine businesses rather than ten.
- The MSCI United Kingdom Index holds 66 stocks and puts 52.48% of its weight in ten of them, with HSBC alone at 11.00%.
- MSCI World Equal Weighted holds the same 1,280 stocks with a 1.40% top ten. Since June 1994 it returned 8.39% a year against the parent index's 9.08%.
- The IMF put the Herfindahl-Hirschman Index above its 95th historical percentile in two of six major equity markets in April 2026.
The top 10 index weight in a 1,280-stock fund is 26.61%
You hold a global index fund because it owns everything. So how much of it sits in the largest ten positions?
MSCI's own factsheet for the MSCI World Index, dated 31 August 2026, gives the number: 26.61%. The index held 1,280 constituents that day. Ten of them carried more than a quarter of the money.
The same document read very differently almost eleven years earlier. On 30 October 2015 the MSCI World Index held 1,642 constituents, and its ten largest came to 9.77%. The methodology didn't change in between. The index still covers "approximately 85% of the free float-adjusted market capitalization in each country", and it still weights holdings by float-adjusted market value. What changed is what the market decided those companies were worth.
So the top 10 index weight rose 16.84 percentage points inside a product people buy precisely so they don't have to pick companies. That's the finding. The rest of this piece is about what it does and doesn't mean.
Index concentration over time isn't a straight line, and 2022 shows why
The chart above plots the top ten weight at ten dates, each read off the MSCI World factsheet published in that month. It reached 10.97% at the end of 2017 and 12.04% at the end of 2018. It was 13.27% in November 2019, then 16.86% at the end of 2020, then 19.81% in November 2021.
Then it fell. At the end of 2022 the top ten came to 15.62%, more than four points below the 2021 reading. Nothing had been fixed. The largest holdings simply fell harder than the rest of the index in a year when growth stocks repriced, which is what a cap-weighted top ten does on the way down. By the end of 2023 it was back to 20.51%, and by the end of 2024 it was 26.30%.
Two things moved at once. The weight climbed, and after 2019 the count of holdings fell: 1,642 names in 2015, 1,650 in November 2019, then 1,395 at the end of 2024 and 1,280 in August 2026. A cap-weighted index doesn't hold a fixed number of stocks. It holds whatever clears the size and liquidity screens, so when the top of the market pulls away, the bottom of it drops out of the index entirely.
The top 10 holdings are not ten companies, and not ten separate bets
On the August 2026 factsheet, Alphabet appears twice: Alphabet A at 2.15% and Alphabet C at 1.69%, or 3.84% between them. Those are two share classes of one business. The list of ten lines is nine companies. NVIDIA is 5.56% of the index on its own, Apple 5.07%, and Microsoft 3.90%.
The skew runs deeper than the top ten. On the same factsheet the largest constituent carries a float-adjusted market capitalisation of US$5,096.7 billion, while the median constituent is US$24.3 billion. The typical company in this index is roughly 210 times smaller than the largest one, and there are more than a thousand of them.
The nine aren't spread across the economy either. Information Technology is 29.81% of the whole index, and the United States is 72.14% of it. A fund sold as 23 developed markets has close to three quarters of its money in one of them. That's a separate measurement from sector concentration, which asks what a capped index actually constrains, but the two compound: the largest names and the largest sector are mostly the same securities.
A wider index isn't the fix, and a narrower one is much worse
If this were a quirk of MSCI World, another index would solve it. On the same date, 31 August 2026, MSCI's own factsheets say otherwise.
- MSCI United Kingdom: 66 constituents, top ten weight 52.48%. HSBC is 11.00%, Shell 7.87% and AstraZeneca 7.58%.
- MSCI Japan: 168 constituents, top ten weight 30.25%.
- MSCI Emerging Markets: 1,178 constituents, top ten weight 37.53%. Taiwan Semiconductor alone is 15.14%, Samsung Electronics 7.20%.
- MSCI World: 1,280 constituents, top ten weight 26.61%.
Emerging markets is the awkward row. It holds nearly as many stocks as MSCI World and is far more concentrated, because one Taiwanese chipmaker is close to a sixth of it. Counting holdings tells you very little on its own. The 66-stock UK index and the 1,178-stock emerging markets index are both more top-heavy than the 1,280-stock world index, and the stock count runs the wrong way in both cases.
The S&P 500 tracker most people own is more concentrated than the world index
State Street's fact sheet for the SPDR S&P 500 ETF Trust, dated 30 June 2026, lists 504 holdings. Its ten largest sum to 36.31%: NVIDIA at 7.50%, Apple at 6.57%, Microsoft at 4.29%, down to Tesla at 1.83%. Information Technology is 38.03% of the fund.
That is 9.7 percentage points more top-heavy than MSCI World, out of 504 holdings rather than 1,280. It matters because the two funds are usually held together. Anyone running a US core alongside a global fund is buying the same nine or ten businesses twice, which is the arithmetic behind fund overlap. The single-stock version of the same problem is set out in our note on NVIDIA index concentration.
The strongest objection: the de-concentrated version of this index has trailed for 32 years
Here is the serious case against everything above. MSCI publishes an equal weighted index built from exactly the same 1,280 constituents as MSCI World, reset to equal weights at each quarterly rebalance. Its largest holding is 0.17% and its top ten is 1.40%. It's the concentration problem solved completely, by construction.
The rebuild is thorough. Its median constituent weight is 0.08%, against 0.03% for the cap-weighted parent, so the typical holding in the equal weighted index carries nearly three times the weight it does next door.
It has also lost. On the equal weighted factsheet, which prints both lines under one column headed 30 June 1994, gross returns to 31 August 2026 ran 8.39% a year for MSCI World Equal Weighted against 9.08% for MSCI World. The risk-adjusted figures on the same sheet cover 1 June 1994 to 31 August 2026 explicitly, and give a Sharpe ratio of 0.42 against 0.47. Over the ten years to August 2026 the return gap was wider: 10.07% against 13.56%. Index turnover in the last twelve months was 31.16% for the equal weighted index and 2.95% for the parent, and turnover is a real cost inside a fund even though the index figures ignore it.
The risk numbers don't rescue it. Maximum drawdown was 59.66% for the equal weighted index against 57.46% for MSCI World, both in the same window from October 2007 to March 2009. Removing the concentration did not remove the crash.
The sign also depends on which index you equal-weight. Our 50-year rebuild of equal weight vs market cap in a US large-cap universe found the equal-weighted version marginally ahead. MSCI World from 1994 gives the opposite answer. Same idea, different universe and different window, opposite result.
The counter-case to the counter-case is narrower than it looks. In 2022, the year the top ten weight fell, the equal weighted index lost 16.38% against the parent's 17.73%. It also beat the parent in 2025, 22.05% against 21.60%. The de-concentrated portfolio does its job in the years the concentrated one hurts. Across the full record it hasn't paid for the years in between.
Regulators measure the same thing and call it a stability risk
The IMF's April 2026 Global Financial Stability Report puts the Herfindahl-Hirschman Index, which it calls "a measure of concentration risk", above "its 95th historical percentile" in "two out of six major equity markets". It reads concentration as a transmission channel, warning that "sell-offs in a small number of firms could cascade into broader market declines".
The Bank of England's July 2026 Financial Stability Report puts the exposure plainly: "The S&P 500 accounts for around half of global equity capitalisation." It adds that within that index, companies it classes as AI-related "now account for around half, up from around a quarter in 2022". Neither institution is making a diversification argument. Both are describing how a repricing in a handful of firms reaches everyone.
That framing is the honest version of the concern. A high top 10 index weight is not evidence that those companies are overvalued. It is evidence that if they are, the index has no mechanism to soften it.
What high concentration has meant for returns before
Vanguard published research on 20 July 2026 using CRSP data running back to 1958. It found the late 1950s looked much like now: "the 10 largest companies accounting for roughly 32% of the market's capitalization". Those same companies "made up just 2% of the market" by December 2025.
That's a 30 percentage point collapse in the weight of a generation's dominant firms, and the market did not collapse with them. Vanguard puts the US market's return since 1958 at "roughly 11%" a year, and "approximately 7%" after inflation. Apple entered at 0.1% of the market in 1980 and was 6.7% in December 2025, behind NVIDIA at 7.6%. Leadership turned over completely and the index kept compounding.
What that record doesn't establish is that any particular concentrated market ends well. It is one country, one sample, and one handover that happened to work. Vanguard's analysis also tracks a cohort forward from a fixed date, which is a different question from whether high concentration predicts weak subsequent returns. On that question this piece has no evidence to offer.
What this evidence can't tell you
Start with the sample. The factsheet series here runs from October 2015 to August 2026, because that is as far back as published MSCI World factsheets survive in the web archive. Almost eleven years is one market regime, not a law, and at annual frequency it contains a single down year.
The dates aren't perfectly regular. Six readings are month-end December, two are November, one is October and the last is August 2026, because those are the months the archived factsheets happened to be captured. The comparison across indices is cleaner, since every MSCI factsheet quoted here is dated 31 August 2026, but the SPDR fund sheet is dated 30 June 2026 and isn't strictly comparable to them.
Index weight also isn't economic exposure. A company earning worldwide revenue from a US listing gives you more geographic spread than the 72.14% US country weight suggests, and a UK-listed miner is not a bet on the UK economy. The count of holdings has the same problem in reverse, which is the argument in how many stocks is diversified.
And every figure above measures weight. None of it measures correlation. Two large holdings that move together are effectively one position, and no factsheet tells you that. A holdings-level view of your own accounts does more here than an index factsheet: LedgerTouch adds up the same company across every fund you hold, which is where the duplication actually appears.
What would change the conclusion
If the top 10 index weight fell back toward its 2022 level of 15.62% without the index losing money, the trend here would look like an episode rather than a structural shift. 2022 proved the number can fall. It fell that year because prices fell, which is not the same thing.
If the equal weighted index started winning, the trade-off flips. Its 8.39% against 9.08% since 1994, at a Sharpe ratio of 0.42 against 0.47, is what makes cap-weighted concentration tolerable. A decade of the reverse would make de-concentration the cheap option rather than the expensive one.
If the two Alphabet lines merged, or a large private company listed, the top ten would rearrange without anything real changing. The Bank of England flags exactly that, noting share offerings that "could materially increase this concentration". A measure that moves on listing decisions is a measure to read carefully.
The figure worth watching isn't 26.61% on its own. It's that number alongside the constituent count, which has fallen at every reading since 2019. A rising top 10 index weight inside a shrinking index is a different animal from a rising one inside a growing index, and only one of the two is happening.