Key takeaways
- A total-market fund, an S&P 500 fund, an S&P 500 growth fund, a technology-sector fund and ARK Innovation, held 20% each, come to 1,533 distinct companies — with 39.3% of the money in the ten largest. The total-market fund on its own holds 1,506 companies with 34.1% in its top ten.
- NVIDIA is 7.3% of the total-market fund. Look through all five funds at once and it is 9.2% of the combined portfolio. Adding four funds raised the single-stock exposure.
- The S&P 500 fund and the S&P 500 growth fund share all 10 of their top-10 names, and 65.8% of their weight. The S&P 500 fund and the total-market fund share all 10, and 92.2%.
- Swap growth, tech and innovation for a value fund, a developed-markets ex-US fund and a small-cap fund, and the same five-fund count produces 3,947 companies with 16.9% in the top ten — an effective 186 equally sized positions against 43 for the first set.
- This isn't a quirk of one index family. MSCI's ACWI index had 23.28% of its weight in ten of its 2,461 constituents at 30 June 2026, and FTSE Russell reports that just 5 securities made up the top 40% of the Russell 1000 Growth Index at 30 December 2025.
Whether five funds diversify you depends entirely on which five
You own five different funds. You want to know whether you own five different things, or the same handful of companies five times over.
The answer is not decided by the number five. It is decided by what the five hold, and you can settle it with published holdings files rather than argument. Take a set that a lot of people end up with by accretion: a broad US total-market fund, an S&P 500 fund, a large-cap growth fund, a technology-sector fund, and a well-known innovation fund, held at 20% each. On State Street's daily holdings files for 23 July 2026 and ARK's for 24 July 2026, that portfolio owns 1,533 distinct companies, and 39.3% of the money sits in the ten biggest of them.
A single total-market fund — one ticker, one line on the statement — owns 1,506 companies with 34.1% in its top ten. The five-fund portfolio is the more concentrated of the two.
Now change three of the five. Keep the total-market fund and the S&P 500 fund, and replace growth, technology and innovation with an S&P 500 value fund, a developed-markets ex-US fund and a small-cap fund. Same five funds. The result is 3,947 companies and 16.9% in the top ten — less than half the concentration of the first set, and less than the single total-market fund on its own.
Same fund count, opposite outcome. The variable is overlap, and overlap is measurable.
Look-through means counting companies, not tickers
The reason a fund count misleads is that risk attaches to companies, not to product wrappers. If NVIDIA falls 30%, it does not matter across how many funds you were holding it. What matters is the total weight of NVIDIA in your portfolio once you look through every fund to its underlying positions — the "look-through" weight.
Computing it is straightforward arithmetic. Multiply each fund's weight in your portfolio by that company's weight inside the fund, then add up the results across all funds. In the five-fund set above, NVIDIA appears in all five: 7.31% of the total-market fund, 7.94% of the S&P 500 fund, 14.67% of the growth fund, 14.10% of the technology fund and 1.79% of the innovation fund, all as reported in the holdings files for 23 and 24 July 2026. Weighted at 20% each, that is 9.2% of the whole portfolio in one company.
The total-market fund alone would have given you 7.3%. The four additional funds added only 27 companies the total-market fund did not already own, and they moved the largest single-stock exposure up by nearly two percentage points. Apple lands at 6.9%, Microsoft at 4.9%, Broadcom at 3.5%. Alphabet, counting both share classes, at 4.4%.
Ticker counts do not capture this and neither does the number of funds. A useful summary statistic is the effective number of stocks: one divided by the sum of every squared portfolio weight, which answers "how many equally sized positions would produce this much concentration?" The five-fund portfolio's 1,533 companies behave like roughly 43 equal positions. The single total-market fund's 1,506 companies behave like roughly 56.
Five funds, five holdings files, and how much of each is the same ten companies
Every figure below is computed from the fund providers' own published holdings, using the equity line items and excluding cash, money-market and futures entries. State Street's four files are dated 23 July 2026; ARK's is dated 24 July 2026. "Weight overlap" is the standard measure: for every company both funds hold, take the smaller of the two weights, and add those up. It answers "what fraction of these two funds is literally the same positions in the same proportions?"
| Fund | Index tracked | Equity holdings | Top 10 as % of fund | Top-10 names shared with the S&P 500 fund | Weight overlap with the S&P 500 fund |
|---|---|---|---|---|---|
| SPDR Portfolio S&P 1500 Composite (SPTM) | S&P Composite 1500 | 1,506 | 34.1% | 10 of 10 | 92.2% |
| SPDR S&P 500 ETF Trust (SPY) | S&P 500 | 503 | 36.9% | — | — |
| SPDR Portfolio S&P 500 Growth (SPYG) | S&P 500 Growth | 147 | 58.3% | 10 of 10 | 65.8% |
| Technology Select Sector SPDR (XLK) | Technology Select Sector Index | 74 | 61.4% | 5 of 10 | 37.5% |
| ARK Innovation (ARKK) | None — actively managed | 46 | 48.1% | 0 of 10 | 13.7% |
Three of the four index funds share their entire top ten with each other: NVIDIA, Apple, Microsoft, Amazon, both Alphabet classes, Broadcom, Meta, Micron and Eli Lilly. The growth fund isn't holding different companies from the S&P 500 fund — it holds 147 of the same 503, in different proportions. The technology fund shares five of ten, and the innovation fund shares none.
The chart plots the same measurement across portfolios rather than funds: 34.1% in the top ten for the total-market fund alone, 36.9% for the S&P 500 fund alone, 39.3% for the five overlapping funds, and 16.9% for the five complementary ones.
Overlap by name and overlap by weight give different answers
This is where the measurement gets slippery, and it's worth being precise about it before drawing conclusions.
Every one of the growth fund's 147 holdings is also held by the S&P 500 fund. By name, the overlap is 100% — there is not a single company in one that is absent from the other. By weight it is 65.8%, because the growth fund holds NVIDIA at 14.7% where the S&P 500 fund holds it at 7.9%. Both statements are true and they support opposite headlines.
The same gap appears at the other end. The technology fund and the innovation fund share only 5 company names out of 74 and 46 respectively, and 9.7% by weight. Their top tens have one company in common, AMD. Whichever number you quote, that pair is doing genuinely different things.
The measure that matters for your risk is neither of those. It is the look-through weight of each company in the portfolio you actually hold, which is the calculation in the previous section — and it depends on how much money you put in each fund, not just which funds you own.
The fund weights change the answer as much as the fund choice
Equal 20% slices are a convenient illustration, not a description of anyone's portfolio. They also flatter the argument, so here's the sensitivity.
Hold the same five funds at 60% total market and 10% in each of the other four, and the top-ten weight falls from 39.3% to 36.3%, with an effective 50 equal positions rather than 43. Still more concentrated than the total-market fund alone, but the gap narrows from about 14 effective positions to about 6. Push the satellite weights lower again and the gap narrows further without closing. At 80% total market and 5% in each of the other four, the top ten is 35.1% against the total-market fund's 34.1%, and 53 effective positions against 56.
That is the honest shape of the finding. Overlapping funds raise concentration in proportion to how much money is in the overlapping funds. A 5% technology sleeve alongside a 95% total-market position is a rounding error on the look-through weights. Five equal slices isn't.
An S&P 500 fund holding 500 companies is still meaningfully diversified
The strongest objection to everything above is that it risks implying index funds are secretly undiversified, which they aren't.
The S&P 500 fund's top-ten weight of 36.9% means 63.1% is spread across 493 other companies. Its effective number of stocks is about 48. That is a long way from the 500 the name suggests, and it is also a long way from a concentrated portfolio. The concentration is not something the fund did — a capitalisation-weighted index fund reports the market's own concentration rather than expressing a view about it. If ten companies are 37% of the S&P 500's value, a fund that held them at 2% each would be making an active bet against them.
US securities law is explicit that this is permitted. Under the Investment Company Act, a "diversified company" must hold at least 75% of total assets in cash and cash items, government securities, shares of other investment companies, and other securities. Only that last category carries the 5%-of-assets and 10%-of-voting-stock caps. The remaining 25% is unconstrained. Two details are easy to miss. The test subclassifies management companies, so it doesn't reach SPY at all: State Street describes the SPDR S&P 500 ETF Trust as "a unit investment trust". And section 5(c) says a fund that qualified when it bought a position keeps its diversified status when that position later appreciates. That's the better explanation for how an index fund ends up carrying a weight above 5%. SPTM held NVIDIA at 7.31% and Apple at 6.83% on 23 July 2026, and State Street's statement of additional information dated 31 October 2025 classifies it as "diversified". ARK Innovation, by contrast, states in its 30 November 2025 prospectus that it's classified as "non-diversified", which it explains means it "may invest a high percentage of its assets in a limited number of issuers".
The label is a legal test, not a risk measurement. Reading it as the latter is a common mistake, and it cuts in the direction of complacency rather than alarm.
Deliberate overlap is a position, not a mistake
The second objection is more important, and it's the reason this piece can't end with a recommendation.
Suppose someone deliberately wants a large-cap growth tilt. Holding an S&P 500 fund and an S&P 500 growth fund produces exactly that: 65.8% of the pair is identical, and the non-identical part is the tilt. The overlap isn't a failure of the portfolio, it is the portfolio. Someone who argues that the two funds are redundant has misread what the second fund is for — its job is to double the weight of names the first fund already holds. A critic of the diworsification framing would point out, correctly, that overlap without intent and overlap with intent look identical on a spreadsheet and are opposite things in practice.
So the distinction that carries the weight isn't overlap versus no overlap. It's whether the resulting look-through weights are the ones you meant to have. A 9.2% position in a single company is a defensible allocation if it was chosen and an accident if it was assembled. The arithmetic can't tell you which. It can only tell you the number, which is the part most portfolios don't have written down anywhere.
There's a third case worth naming: overlap that shows up despite deliberate diversification. The developed-markets ex-US fund in the second portfolio shares no top-ten names with any US fund, and the small-cap fund's largest position is 0.56% of its assets. Those two genuinely reduce the look-through concentration, and the numbers show it. Diversifying by geography and size did what diversifying by fund count did not.
Global and non-US index families show the same shape
None of this is an artefact of one index provider. MSCI's ACWI index — 2,461 constituents across 23 developed and 24 emerging markets, covering roughly 85% of the global investable equity opportunity set — had 23.28% of its weight in its ten largest constituents at 30 June 2026, with the United States at 63.63% of the index and information technology at 32.09% of it. Nine of those ten largest were US-listed; the exception was Taiwan Semiconductor.
FTSE Russell's own factsheet for the Russell 1000 Growth Index, prepared for the US tax code's qualified-index tests and reporting data as of 30 December 2025, discloses that 5 securities made up the top 40% of index weight and that the largest single security, NVIDIA, was 12.224% after capping. A differently built growth index, from a different provider, on an earlier date, arrives at the same structural place.
Those two figures come from different as-of dates than the fund holdings above and can't be added to them. They are corroboration of the pattern, not inputs to the portfolio arithmetic.
What this measurement can't tell you
Holdings change, and these are a snapshot. Every fund figure here's a single day: 23 July 2026 for the State Street funds, 24 July 2026 for ARK Innovation. An actively managed fund can look completely different a quarter later, and index funds shift with the market's own weights. Repeat this exercise in six months and the percentages are unlikely to match; whether the structure has changed is the open question.
Two of the funds are also not quite what their shorthand suggests. The S&P Composite 1500 covers approximately 90% of the investable US equity market, not all of it — micro-caps sit outside. And the innovation fund's holdings file includes private positions, with Space Exploration Technologies at 4.66% and an OpenAI series at 3.03% on 24 July 2026, which aren't marked in a public market at all.
The effective-number-of-stocks figure treats every company as equally risky and ignores that many of these names move together. Four of the ten largest positions in the combined portfolio are semiconductor businesses — NVIDIA, Broadcom, AMD and Micron — which the concentration arithmetic counts as four separate holdings but which a correlation matrix would count as rather fewer. The look-through weight is a floor on the concentration you carry, not a ceiling.
Finally, the weight overlap measure says nothing about future behaviour. Two funds with 92% overlap are close to the same portfolio, but two funds with 14% overlap aren't therefore uncorrelated.
What would change the conclusion
If mega-cap weights compressed. The entire effect runs through the size of the largest positions. With NVIDIA at 7.3% of the US market, stacking growth and technology funds on top of it moves the number a lot. If the top ten were 15% of the S&P 500 rather than 37%, the same five funds would produce a look-through portfolio barely distinguishable from the total-market fund, and this arithmetic would be a curiosity.
If the added funds tilt in different directions. The second portfolio above already demonstrates this: value, international and small-cap exposure pulled the top-ten weight down to 16.9% and the effective position count up to 186. Five funds aren't the problem. Five funds that all express the same tilt are a different thing from five funds that express five tilts.
If the concentration was chosen. A reader who wanted 9.2% in one company and 39.3% in ten isn't diworsified — they are positioned, and this piece describes their portfolio accurately without criticising it. The finding is about knowing the number, not about what the number ought to be.
The figure worth having is the look-through weight of your largest company across every fund you own, alongside the weight you would have had from a single broad fund. Fund providers publish the holdings files that make that computable, and portfolio trackers including LedgerTouch do the multiplication for you. What neither can do is tell you which of the two numbers you wanted.
The related question — whether funds that diversify by geography earn their place, and what happens to single-stock index weight when one company dominates — is where this arithmetic leads next.