SMH or SOXX is the most common question in semiconductor ETFs, and the most common answer — hold both — is the one the filings argue against. The two funds share 23 companies and 74% of their weight. Hold them together in equal size and the combination behaves like 18.0 equal holdings, fewer than SOXX's 20.4 on its own. The pair does not diversify the semiconductor position; it sizes it twice, with extra NVIDIA.
What separates them is one rule. SOXX caps every position, so nothing is above 8.5%. SMH does not cap as hard, so NVIDIA is 17.6% and TSMC 9.3%. Everything else follows from that — including why SMH has grown more concentrated since 2019 while SOXX has not moved.
The figures
- Both funds filed for the period ending 2026-06-30 (SMH on 2026-08-26, SOXX on 2026-08-25). SMH: 25 stocks, $77 billion, top ten 70.3%, NVIDIA 17.6%, TSMC 9.3% — it behaves like 13.9 equal holdings. SOXX: 32 stocks, $48 billion, top ten 61.7%, largest position 8.5% — it behaves like 20.4.
- Overlap by weight: 74.0%, across 23 shared companies. Held in equal size the pair behaves like 18.0 holdings, 2.4 fewer than SOXX alone. Adding SMH to SOXX concentrates the portfolio; adding SOXX to SMH loosens it, but not to where SOXX already is.
- What only SMH holds: Cadence and Synopsys (chip-design software, 3.9% together). What only SOXX holds: Credo, Entegris, ASE, MACOM, United Microelectronics and three smaller names (8.1% together). Everything else is the same companies at different weights — NVIDIA is 17.6% of SMH and 6.8% of SOXX; Micron and AMD are larger in SOXX.
- SMH has become more concentrated; SOXX has not moved. SMH behaved like 16.5 holdings in 2019, 11.8 in March 2024 and 13.9 now, as its largest position went from 12.9% to 20.6% to 17.6%. SOXX's effective count has stayed near 20 for seven years because its index caps weights.
- 82% of SMH and 84% of SOXX is US-domiciled; the rest is Taiwan (TSMC, ASE, UMC) and the Netherlands (ASML). This describes filings on a date; it is not a view on semiconductors or on either fund.
Data as of 2026-06-30 (filing period) · received by the SEC 2026-08-25/26 · written 2026-09-25. The next period (2026-09-30) is due by the end of November and will be re-measured here.
Side by side
Source: SMH Form N-PORT, filed 2026-08-26 and SOXX Form N-PORT, filed 2026-08-25, rows in the point-in-time files; the full pair on the comparison page.
| SMH | SOXX | |
|---|---|---|
| Positions | 25 | 32 |
| Net assets | $77.4B | $48.4B |
| Largest position | NVIDIA 17.6% | Micron 8.5% |
| Top ten | 70.3% | 61.7% |
| Behaves like | 13.9 | 20.4 |
| US / Taiwan / Netherlands | 82 / 9 / 7 | 84 / 6 / 6 |
The difference is a rule. SOXX's index caps single positions (no name is above 8.5% in this filing, and across 28 filings since 2019 the largest has ranged from 7.9% to 11.5%); SMH's index lets the largest names run to a higher cap, so NVIDIA sits at 17.6% and TSMC at 9.3%. Same industry, 23 of the same companies, and one fund is half again as concentrated as the other because of how the weights are set.
What each holds that the other does not
| Only in SMH | Weight | Only in SOXX | Weight | |
|---|---|---|---|---|
| Cadence Design Systems | 2.2% | Credo Technology | 2.0% | |
| Synopsys | 1.8% | Entegris | 1.2% | |
| ASE Technology | 1.2% | |||
| MACOM Technology | 1.2% | |||
| United Microelectronics | 1.0% | |||
| three others | 1.5% | |||
| Total | 3.9% | Total | 8.1% |
SMH's two exclusives are software companies that design chips rather than make them. SOXX's are smaller manufacturers, packagers and materials suppliers. Neither list is large: 96% of SMH and 92% of SOXX is the shared 23.
The shared 23, at their two weights
| Company | SMH | SOXX |
|---|---|---|
| NVIDIA | 17.6% | 6.8% |
| Micron Technology | 5.7% | 8.5% |
| Advanced Micro Devices | 5.6% | 8.1% |
| Intel | 5.2% | 6.3% |
| Broadcom | 5.3% | 6.1% |
| Applied Materials | 5.7% | 5.8% |
| KLA | 5.6% | 5.6% |
| Lam Research | 5.3% | 4.9% |
Overlap counts, for each shared company, the lesser of the two weights; that is where the 74% comes from. The gap between the funds is almost entirely NVIDIA: 17.6 against 6.8. Strip that one difference and the two portfolios are nearly the same list at nearly the same weights.
Owning both
| Held | Behaves like |
|---|---|
| SOXX alone | 20.4 |
| SMH alone | 13.9 |
| SMH + SOXX, equal size | 18.0 |
A holder of SOXX who adds SMH ends up more concentrated than before, because the addition is mostly the same companies with more NVIDIA. A holder of SMH who adds SOXX ends up less concentrated, but still short of simply holding SOXX. Two semiconductor funds are not diversification across semiconductors; they are one semiconductor position sized twice.
Seven years
| Filing period | SMH: behaves like | Largest | SOXX: behaves like | Largest | |
|---|---|---|---|---|---|
| 2019-09 | 16.5 | 12.9% | 19.8 | 8.4% | |
| 2021-03 | 16.6 | 15.2% | 19.5 | 8.6% | |
| 2022-09 | 18.9 | 11.3% | 21.1 | 8.8% | |
| 2024-03 | 11.8 | 20.6% | 20.6 | 8.8% | |
| 2025-09 | 12.5 | 19.3% | 21.4 | 7.9% | |
| 2026-06 | 13.9 | 17.6% | 20.4 | 8.5% |
SMH's concentration follows its largest position, which is NVIDIA's market value; the fund was at its tightest in early 2024. SOXX's cap makes its figure indifferent to which company is largest. The two funds' different histories are the same rule difference, seen over time.
What the two actually returned, as they filed it
Not a price feed and not our calculation: the monthly total returns each fund reported in Item B.5 of its own N-PORT, compounded.
| Whole series | Months | Last 12 months | Best month | Worst month | |
|---|---|---|---|---|---|
| SMH | +1,154.0% | 84 | +136.1% | +32.2% | −16.6% |
| SOXX | +931.1% | 84 | +169.6% | +40.3% | −17.7% |
Two things in that table point opposite ways, and both are worth holding at once.
Over seven years the more concentrated fund returned more: SMH's 1,154% against SOXX's 931%, a gap of 223 points that came from letting NVIDIA run toward a fifth of the fund while SOXX's cap kept trimming it back.
Over the last twelve months on record the less concentrated fund returned more: SOXX +169.6% against SMH +136.1%. SOXX's best single filed month, +40.3%, is eight points better than SMH's best.
So the cap cost SOXX the long run and paid it back recently. Neither figure is a reason to hold either fund, and the honest summary is narrower than either number suggests: a concentration measure describes the shape of a portfolio and says nothing about which way that shape will pay. This site publishes it because it is in the filings, not because it forecasts.
Both series agree with an independently computed price series, so they are published rather than withheld; the funds where our two readings disagree are named here.
One of them lends almost nothing, to almost nobody
A line from Item B.4.a, where a fund names every firm borrowing its stock:
| On loan, % of fund | Borrowers named | Largest borrower's share | |
|---|---|---|---|
| SMH | 0.11% | 3 | 97.0% |
| SOXX | 1.28% | 8 | 39.6% |
SMH lends a ninth of what SOXX does, and effectively all of its small loan book sits with one counterparty. At 0.11% of the fund the amounts are too small to matter to a holder — the point is the contrast, because it is one more way two funds holding the same 23 companies are not the same instrument. The whole lending dataset, and a fund where this does matter.
What this can and cannot tell you
- As of 2026-06-30. Both funds' weights move with prices daily and reset at their indexes' rebalances.
- Concentration is about weights, not about which companies; 13.9 holdings in chip designers and 13.9 in anything else carry different risks this measure cannot see.
- Country is the issuer's country as filed: TSMC and ASE as Taiwan, ASML as the Netherlands, all held through US-listed shares.
- Nothing here is a view on semiconductors or on which fund to hold.
Data: SMH vs SOXX · SMH · SOXX · overlap.json · position files for SMH and SOXX.