What Is Concentration Risk? The S&P 500's 10 Largest Hold 39.48% (Oct 2026)

Michael Whitman · Trading since 2017, full-time since 2020 · YouTube creator, 137K subscribers · Published 2026-09-14 · Updated 2026-11-02 · Reviewed 2026-11-01

Concentration risk is the danger that one company, sector or shared driver decides most of a portfolio’s result, however many holdings it lists. In the S&P 500 fund SPY on 23 Oct 2026, ten of 500 companies made up 39.48% of the stock held.

How it works

A concentrated exposure is one where a single event determines the result. One issuer defaulting, one sector repricing, one currency moving.

Counting positions measures nothing. Twenty holdings that respond to the same driver behave as one position with extra transaction costs attached.

The right measure is how the holdings move together, and how the money is split between them. If they rise and fall in step, the portfolio has one source of variation regardless of how it is labeled. And if most of the money sits in a few lines, the other lines barely count.

Where it hides

Inside index funds. A market-capitalization index weights each company by its market value, so whatever has risen most takes a larger share. A fund holding 500 companies can end up with a large part of its value in a handful of them, which the SPY figures below show.

In employment. Somebody holding company stock, with a pension in that company, working for that company, has three claims on one outcome, and only one of them shows up in an investment account.

And in shared factors. Holdings across different sectors and countries can still share one sensitivity — to interest rates, to energy prices, to one currency — that no sector breakdown displays.

A worked example

Take $10,000 spread across SPY’s stocks in the proportions the fund held on 23 Oct 2026. That is 500 companies, and on a statement it looks as diversified as anything can.

Now follow the money. $778 sits in Nvidia, $773 in Apple and $593 in Microsoft. The ten largest companies take $3,948. The 250 smallest, half of the list, take $862 between them.

So a 10% fall in the ten largest alone would cost about $395, while the same fall spread across the 250 smallest would cost about $86. The statement shows 500 lines; the result is decided mostly by ten.

And the ten are not independent of each other. Whatever they share, a customer base, a sensitivity to rates, a theme investors buy together, is the driver to look for next.

Now the single-name version, with real prices. Meta’s adjusted close peaked on 7 Sep 2021 and bottomed on 3 Nov 2022. Over those exact dates, $10,000 in Meta fell to about $2,326 (-76.74%), while $10,000 in SPY fell to about $8,369 (-16.31%). Nasdaq’s own closes put Meta’s 2021-22 high and low on the same two days, at $382.18 and $88.91, the same 76.74% fall.

Then mix them. An account that held $9,000 in SPY and $1,000 in Meta ended at about $7,765. One that had let Meta grow to half the account, $5,000 and $5,000, ended at about $5,348.

The difference between those two accounts is concentration and nothing else. Same two holdings, same dates, same prices; only the split changed. Meta’s adjusted close did not get back to its 2021 peak until 19 Jan 2024, so the half-and-half holder waited more than two years to be whole on that line.

How correlation misleads you

Measured correlation is an average over a period. Two holdings that averaged 0.3 over five years may have run much closer together during the three months anybody cared about.

And correlations tend to rise in stress. A shock that moves everything moves everything, so the diversification measured in calm conditions can be smaller when it is needed.

Which means a correlation matrix can understate concentration. It is built from history that is mostly calm, and applied to a future in which the interesting moments are not.

The practical response is not a better model. It is to look at what the holdings actually depend on — the same customer, the same rate, the same commodity, the same regulator — and to treat that as the real count, independent of any statistic.

The original data

The sample: SPY’s full holdings file from State Street, holdings as of 23 Oct 2026. It lists 503 stock lines from 500 companies; Alphabet, Fox and News Corp each appear twice (two share classes) and are counted once here. A cash line (0.30% of the fund) and one contingent value right are left out, and each company’s weight is its share of the fund’s stocks.

Half of the fund’s stock sat in 21 companies. The smallest 236 companies together weighed 7.72%, less than Nvidia’s 7.78% on its own. The ten largest were Nvidia, Apple, Microsoft, Alphabet, Amazon, Broadcom, Meta, Tesla, Eli Lilly and Berkshire Hathaway.

One number sums it up: 42.30. One divided by the sum of each company’s squared weight gives the count of equal-sized holdings that would be exactly as concentrated; it is one divided by the Herfindahl-Hirschman index (HHI) that regulators use for market concentration, with the weights written as fractions rather than percentages. By that measure SPY’s 500 companies carried the concentration of about 42 equal positions. Every company’s weight and running total is in the holdings file.

Checked against Nasdaq’s market capitalizations (at the 23 Oct 2026 close, the same day as the holdings, with every share counted rather than the index’s float-adjusted count; Brown-Forman has no value there, so 499 companies): the same ten companies were the largest, with 39.83% of the total, 20 companies reached half, and the measure came to 43.74. Two different sources, the same picture.

Even the winners fell hard

One more test, from daily prices this time. We took SPY and the same ten companies (Alphabet’s class A shares, Berkshire’s class B) and measured each one’s worst fall from a closing high to a later closing low, on dividend-adjusted closes from 2 Jan 2015 to 30 Oct 2026. That is 2,975 trading sessions for each.

Nine of the ten fell further than SPY’s 33.72%, and the median of the ten was 46.31%. Meta lost 76.74%, Tesla 73.63% and Nvidia 66.34%. Only Berkshire Hathaway, at 29.57%, held up better than the fund.

This sample is tilted toward survivors, which is worth saying plainly. These are the companies that ended up largest, picked with hindsight, and every one of them was worth more on 30 Oct 2026 than on 2 Jan 2015.

Even so, holding any one of them alone meant sitting through a fall of between about 30% and 77% at some point. Companies that shrank out of the top ten are not in the sample at all. Every peak, low and recovery date is in the drawdown file.

The broad market itself shows how deep one shared driver can cut. The S&P 500 fell 56.8% from its 9 October 2007 close to 9 March 2009, and it did not close above that peak again until 28 March 2013.

Why it persists when everybody knows about it

Concentration is often how wealth was built. A concentrated position that worked is the reason many large individual holdings exist, and the instinct that built it resists dismantling it.

Tax makes reduction expensive. A large gain means selling triggers a bill, and the deferral argument for holding is real even when the risk argument points the other way.

And familiarity feels like information. People concentrate in what they know — their employer, their industry, their country — and mistake that knowledge for an edge that justifies the exposure.

How to actually measure it

Look through to the drivers. List what each holding depends on - a rate, a commodity, a currency, one customer, one regulator - and count the distinct drivers rather than the positions.

Then count by weight, not by line. The largest ten as a share of the total, and one divided by the sum of squared weights, take a minute in a spreadsheet and say more than the number of holdings.

Ask what single event costs the most. A one-line summary of the worst plausible outcome is often more useful than any correlation matrix, and it requires no model at all.

And set a limit before you need one. A rule capping any single driver at a stated share is crude and mechanical, which is the point: it is decided before the position has grown.

How this connects

The industry-level version is the concentration ratio, which counts how much of a market a few firms control; a portfolio heavy in one such industry inherits that structure.

Hidden drivers usually have names. A shared sensitivity to rates is bond duration in disguise, a shared sensitivity to one currency is currency risk, and a cluster in one jurisdiction is country risk.

Concentration is the part of risk that could have been avoided. In standard portfolio theory, market risk remains whatever you hold, while risk tied to one company can be diversified away, which is why that theory pays no extra expected return for carrying it.

Lenders meet the same problem as credit risk. A loan book full of one industry fails together, and a trading book facing one bank holds counterparty risk in the same shape.

Unwinding it has tools. Selling in stages spreads capital gains tax across years, tax loss harvesting elsewhere can offset part of the bill, and a covered call or a put option can reduce exposure while a sale is phased in.

Some hold it on purpose. Charlie Munger argued for a few well-understood positions and accepted concentration as the price.

When it fails

The characteristic failure is a diversified-looking portfolio with one hidden driver. The holdings are numerous, the sectors are several, the geographies are spread, and every report confirms it. Then one variable moves — a rate, a currency, a single large customer common to several holdings — and the whole portfolio moves as one.

A second failure is counting positions as diversification. 500 lines in SPY carried the weight of about 42.

A third is ignoring the employer concentration, which sits outside every investment report.

A fourth is assuming an index fund is diversified by definition. It is diversified by rule, and a market-value rule hands the most weight to whatever has grown largest.

And a fifth is relying on a correlation figure measured in calm conditions, which is the number least likely to hold when it matters.

Questions people ask

What is concentration risk in simple terms?

Too much of the result depending on one thing: one company, one sector, one country or one shared driver such as interest rates. The number of holdings does not tell you whether you have it; the weights and the drivers do.

Is an S&P 500 index fund concentrated?

More than its 500 names suggest. In SPY on 23 Oct 2026, the ten largest companies held 39.48% of the stock, 21 companies held half, and by weight the fund was as concentrated as about 42 equal positions.

How do you measure concentration risk?

Add up the weight of the largest holdings, and work out one divided by the sum of squared weights, which gives the equivalent number of equal positions. Then list the drivers behind the biggest weights and count those instead of the lines.

How do you manage concentration risk?

Set a cap on any one company, sector or driver before a position grows, then rebalance back to it on a schedule. Look inside each fund for holdings you already own elsewhere. The cost of skipping this is visible in real prices: from September 2021 to November 2022, a half-Meta, half-SPY account lost 46.52%, while SPY alone lost 16.31%.

How many stocks make a portfolio diversified?

There is no count that settles it, because the weights and the shared drivers matter more than the number. SPY’s 500 companies measured like 42 equal positions, and twenty holdings tied to one rate can behave like one.

Related

Systemic risk covers concentration at the level of a whole financial system. Risk management covers the wider practice this sits inside. And position sizing covers the control that actually limits it.

The industry version of the same measurement, the share held by the biggest four firms, is the concentration ratio.

This page is educational, not financial advice. Test every idea on your own charts before risking money.