Part 06: Satellites, Thematic Exposure and Concentration

US Markets and Portfolio Construction, unified nine-part series. Research edition: 15 September 2026.

Thematic exposures can be useful only when their economic rationale, overlap and loss budget are visible. Semiconductor funds provide a demanding test case.

Series scope

A practical research question is whether a semiconductor ETF can offer the low-cost characteristics associated with broad building blocks such as SPYM and QQQM without concealing a different concentration profile. SOXQ is a natural candidate because its stated expense ratio is lower than several larger semiconductor funds. That observation is useful but incomplete. A sector ETF is not selected by price alone. Its index methodology determines what the investor owns, how concentrated the exposure becomes, and how the sleeve interacts with the broad U.S. core established earlier in this series.

As of September 2026, Invesco reports a 0.19% total expense ratio for SOXQ [1]. BlackRock reports 0.33% for SOXX [2]. VanEck reports 0.35% for SMH [3], and State Street reports 0.35% for XSD [4]. Those differences are verifiable. But the funds track different semiconductor indexes, use different weighting rules and hold different numbers of securities. A 16-basis-point fee advantage cannot be evaluated separately from those design choices.

This article therefore reframes the decision as: What semiconductor exposure is the portfolio trying to buy, and which ETF delivers that exposure with acceptable concentration, cost, liquidity and overlap? The calculations are educational examples rather than return forecasts.

1. Why a semiconductor ETF belongs in the satellite, not the core

Semiconductors sit upstream of a remarkable range of economic activity: cloud computing, artificial intelligence, automobiles, communications equipment, industrial automation, consumer devices and data centers. That makes the industry strategically important, but importance is not the same as diversification.

A broad S&P 500 ETF already owns major semiconductor companies. A Nasdaq-100 ETF also holds many large chip and semiconductor-equipment firms when they meet index eligibility. Adding a semiconductor ETF therefore overweights an industry already embedded in the core. It should be described as a sector tilt.

This distinction matters because concentration changes the portfolio’s failure modes. A sector sleeve can outperform strongly when earnings, capital spending and valuations move in its favor, yet suffer sharper drawdowns when the semiconductor cycle turns. State Street explicitly warns that sector-focused funds can be more volatile than broadly diversified funds [4]. Invesco likewise notes that SOXQ is non-diversified and that concentration in a specific industry can increase risk [1].

In portfolio language, the semiconductor allocation is not a separate asset class. It is a high-beta, high-conviction equity satellite layered onto existing U.S. equity beta.

2. The industry itself is cyclical

The long-term technological story does not eliminate cyclicality. Academic work on the semiconductor industry has documented industry-specific cycles involving demand, inventories, capacity investment and capital expenditure [5]. Semiconductor manufacturing also requires large, lumpy investment decisions; when supply and demand expectations are wrong, the adjustment can be painful.

Current industry data reinforce the scale of the cycle. The Semiconductor Industry Association and World Semiconductor Trade Statistics publish recurring market data and forecasts rather than assuming a smooth growth line [6], [7]. Forecasts can change materially as memory pricing, AI demand, geopolitical restrictions, capital spending and end-market demand evolve. These sources are useful context, but they should not be converted into a straight-line expected-return assumption for an ETF.

A sector allocation therefore needs a maximum weight and a rebalance rule. The thesis may be secular, but the price path can be cyclical.

3. SOXQ: low cost, concentrated benchmark

Invesco states that SOXQ tracks the PHLX Semiconductor Sector Index and generally invests at least 90% of total assets in securities comprising that index [1]. The fund’s current expense ratio is 0.19%. Invesco describes the underlying index as including 30 of the largest U.S.-listed companies engaged in the semiconductor business, with scheduled reconstitution and rebalancing [1].

Nasdaq’s index overview describes the PHLX Semiconductor Sector Index (SOX) as a modified market-cap-weighted index of 30 semiconductor companies involved in design, distribution, manufacture and sale [8]. The key phrase is modified market-cap weighted. SOXQ is not equal-weighted, and it is not automatically a broad representation of every company in the semiconductor ecosystem.

SOXQ’s strengths for a long-horizon investor are straightforward:

• currently the lowest stated expense ratio among the four funds reviewed here;
• a transparent, established 30-company benchmark;
• straightforward sector purity; and
• an index structure that gives larger businesses larger influence, subject to index methodology.

Its limitations follow from the same design. Thirty holdings is not a large universe. Modified market-cap weighting can produce significant exposure to the largest constituents. The fund is therefore a concentrated sector bet even if no single stock appears extreme in isolation.

4. SOXX: a different benchmark with a larger trading ecosystem

SOXX is the iShares Semiconductor ETF. BlackRock’s current product page reports an expense ratio of 0.33% and identifies its benchmark as the NYSE Semiconductor Index [2]. The fund is one of the largest dedicated semiconductor ETFs and trades with substantial volume, which can matter to investors who use options or execute larger tactical positions.

The critical point is that SOXX does not simply charge more for the exact same portfolio as SOXQ. It follows a different index. Differences in eligibility, weighting, capping, reconstitution and constituent definitions can create different exposures and tracking paths.

For a buy-and-hold investor making small periodic purchases, a 14-basis-point annual fee difference versus SOXQ is meaningful over long horizons. For a trader who values spreads, depth or derivative liquidity, total implementation cost may point in another direction. The correct comparison therefore includes both holding cost and trading utility.

BlackRock also announced a forward stock split for SOXX in 2026 [2]. A split changes share count and nominal share price, not the economic value of the position. It is therefore not a reason by itself to prefer or avoid the fund. This is a useful example of filtering economically irrelevant headlines from portfolio decisions.

5. SMH: fewer names, global U.S.-listed exposure and concentration

VanEck’s Semiconductor ETF, SMH, tracks the MVIS US Listed Semiconductor 25 Index and currently reports a 0.35% expense ratio [3]. The index name contains an important detail: U.S.-listed is not the same as U.S.-domiciled. The exposure can include foreign semiconductor companies whose securities trade in the United States, broadening the industry opportunity set in a way that may differ from other benchmarks.

SMH is also built around a smaller number of names than XSD and can be more concentrated in mega-cap semiconductor leaders. That concentration can be attractive to an investor who specifically wants the industry’s largest global franchises to dominate the sleeve. It can be undesirable to an investor who believes the largest names are already heavily represented in SPYM and QQQM.

This illustrates why “semiconductor ETF” is too broad a category for allocation decisions. Two funds can both be highly correlated with the semiconductor cycle while differing materially in single-name concentration, foreign issuer exposure and sensitivity to the largest AI-related companies.

6. XSD: modified equal weighting changes the bet

State Street’s XSD tracks the S&P Semiconductor Select Industry Index and currently reports a 0.35% gross expense ratio [4]. Its distinguishing feature is the index’s modified equal-weight approach. State Street notes that the structure provides exposure across large-, mid- and small-cap semiconductor companies [4].

Equal or modified equal weighting changes portfolio behavior. Compared with market-cap weighting, it reduces the influence of the biggest firms and increases the relative contribution of smaller constituents. That can reduce single-name concentration, but it can also increase exposure to smaller, more volatile companies and require more rebalancing turnover.

XSD is therefore not the “diversified version” in a risk-free sense. It diversifies company weight, not necessarily economic factor exposure. During a broad semiconductor downturn, dozens of smaller chip names can fall together. Equal weighting may reduce the damage from one mega-cap disappointment while increasing sensitivity to industry breadth and smaller-company financing conditions.

7. Side-by-side design comparison

The table deliberately avoids declaring a universal winner. SOXQ is the clearest cost analogue to SPYM and QQQM, but its benchmark is not economically identical to SOXX, SMH or XSD.

8. Fee drag calculation

Assume $50,000 is invested for 20 years. Suppose, purely to isolate fees, that each underlying benchmark earns the same 9.00% gross annual return before fund expenses. This assumption is unrealistic as a performance forecast because the benchmarks differ; it is used only to quantify fee drag.

Calculation table 1: fee-only sensitivity

The exact dollar values vary with the assumed return and holding period. The inference is narrower: persistent fee differences compound. But because the indexes are not identical, realized benchmark differences could easily swamp the fee gap in either direction. Cost should break ties between suitable exposures; it should not manufacture equivalence between unsuitable ones.

9. Look-through overlap with SPYM and QQQM

The biggest portfolio-construction issue is overlap. Suppose a $100,000 portfolio is 60% SPYM, 15% QQQM and 10% SOXQ, with 15% elsewhere. If a hypothetical semiconductor company is 3% of SPYM, 8% of QQQM and 10% of SOXQ, its portfolio weight is:

The investor may look at SOXQ and see only a 10% satellite, yet the company’s total portfolio exposure is four times the 1% contribution coming from the satellite alone. The same principle applies at the sector level.

The workbook therefore treats overlap as look-through exposure, not as a checkbox that asks whether two funds share any names. The practical workflow is to download current holdings, map common tickers, multiply fund weights by portfolio weights, and aggregate by security and sector. Because ETF holdings change, this process should be refreshed rather than hard-coded into an article.

10. Concentration metrics: HHI and effective number of holdings

A simple way to compare concentration is the Herfindahl-Hirschman Index:

The reciprocal gives an intuitive “effective number of equal-sized holdings”:

If a fund has 25 holdings but a few names dominate, its effective number can be far lower than 25. Conversely, a modified equal-weight portfolio can have an effective number closer to its actual constituent count.

Calculation table 2: hypothetical concentration example

This does not use actual ETF holdings. It demonstrates why holding count alone can be misleading. A concentrated 25-stock ETF and a balanced 25-stock ETF do not carry the same single-name risk.

11. Sector risk is more than volatility

Volatility is useful but incomplete. Semiconductor satellites can be exposed to several non-independent risks:

Demand-cycle risk. End-market demand can weaken rapidly after periods of shortage or over-ordering.

Capital-intensity risk. Fabrication capacity requires enormous investment and long lead times.

Technology-transition risk. Competitive advantage can change when process nodes, architectures, accelerators or memory technologies shift.

Customer concentration. Some companies depend heavily on a few major buyers or hyperscalers.

Geopolitical and supply-chain risk. Advanced semiconductor supply chains span design software, equipment, foundries, memory, packaging and critical materials across jurisdictions.

Valuation risk. A strong secular growth story can still produce poor investment returns if the starting valuation embeds even stronger expectations.

A sector ETF reduces company-specific failure risk but cannot remove the industry’s shared economic risks.

12. Stress testing the satellite weight

Instead of choosing 10% because round numbers feel sensible, derive a ceiling from loss tolerance. Let \(L_p\) be the maximum portfolio loss the investor is willing to attribute to the semiconductor satellite in a severe sector drawdown and \(D_s\) be the assumed satellite drawdown.

If the maximum acceptable portfolio damage is 5% and the stress assumes the semiconductor ETF falls 50%, the maximum satellite weight is 10%.

Calculation table 3: satellite loss budget

This method does not predict the next drawdown. It asks whether the portfolio remains survivable if a plausible severe scenario occurs.

13. Liquidity versus holding cost

Long-term ETF investors and active options traders can rationally choose different wrappers for the same broad theme. Holding cost is primarily driven by expense ratio and tracking quality. Trading cost depends on spread, depth, average volume, order size and derivative liquidity.

SOXX’s larger trading ecosystem may be relevant to an options-oriented investor. SOXQ’s lower fee may matter more to a recurring long-term accumulator. SMH may be preferred for a particular large-company/global exposure. XSD may be preferred for a more balanced company-weight profile.

This is exactly the same principle discussed in Part 1 when comparing SPYM with more heavily traded S&P 500 vehicles: the cheapest long-term wrapper does not have to be the best tactical instrument. Product selection should reflect the use case.

14. Dividend-focused satellites: define the job first

A dividend sleeve can mean at least three different things. It can seek companies with a record of dividend growth, a high current distribution rate, or a quality screen that includes dividends among several indicators. Those choices can overlap, but they should not be treated as interchangeable. A high yield may reflect a mature cash-generative business, a temporarily depressed share price, or a distribution that requires separate sustainability analysis.

SCHD is an example of a dividend-quality implementation whose methodology and holdings should be reviewed from the issuer's current disclosures [10]. A dividend satellite belongs beside the core only when the investor can state why its cash-flow, quality or factor exposure is useful, how much overlap it has with the core, and what maximum allocation applies. Dividend yield is a component of return, not a substitute for total-return and concentration analysis.

15. A practical selection scorecard

The following scorecard avoids false precision by separating criteria instead of collapsing everything into one historical-return ranking.

For the “QQQM/SPYM-like cost efficiency” question, SOXQ scores well on fee. For a broader portfolio decision, all nine rows must be considered.

16. An illustrative allocation decision

Suppose the strategic policy is 60% SPYM, 15% QQQM, 10% semiconductor ETF and 15% tactical/reserve. The semiconductor sleeve has one job: add a bounded industry overweight. It is not expected to provide crash protection, income stability or geographic diversification.

Under that policy, SOXQ can be a reasonable implementation candidate when the investor prioritizes low annual fund expenses and accepts the PHLX index design. XSD may better fit an investor who wants less mega-cap concentration. SMH may better fit an investor who deliberately wants the largest U.S.-listed global semiconductor leaders. SOXX may be attractive when its scale, liquidity or derivative ecosystem materially improves execution.

Notice what is absent from the decision: “Which one returned the most last year?” The answer to that question can change with the measurement window and often rewards whatever concentration happened to win recently.

17. Rebalancing a volatile satellite

A sector sleeve should have a target and a band. If the policy target is 10% and the allowed range is 8%–12%, a strong rally that lifts the sleeve to 14% triggers a rebalance under a threshold policy. A severe decline that reduces it to 7% triggers a review and potentially a purchase, but only if the strategic thesis remains intact.

The distinction between rebalancing and averaging down is governance. Rebalancing follows a pre-written policy. Averaging down can be an improvised response to loss. If fundamentals or benchmark methodology change enough to invalidate the thesis, the correct action may be to change the policy rather than mechanically restore the old weight.

Contribution-first rebalancing can reduce turnover: direct new cash to underweight sleeves before selling overweight funds. The workbook calculates target gaps and shows both full-trade and contribution-first paths.

18. What would change the choice?

ETF selection should be revisited when one of the variables that justified the choice changes. Examples include:

• material expense-ratio changes;
• benchmark methodology changes;
• a merger, closure or ticker change;
• persistent tracking problems;
• materially worse trading conditions;
• a change in the portfolio’s options use;
• a different target for mega-cap concentration; or
• the removal of the semiconductor tilt from the policy itself.

Fund due diligence is not a one-time event. A change in fees, methodology, corporate actions or trading conditions can alter a previously sound implementation decision.

19. Takeaways

SOXQ is currently the strongest match to the low-cost characteristic associated with SPYM and QQQM, with a stated expense ratio of 0.19% [1]. But it is not a cheap clone of SOXX, SMH or XSD. SOXQ tracks a 30-company PHLX benchmark; SOXX tracks a NYSE semiconductor index; SMH tracks a 25-name U.S.-listed global semiconductor index; and XSD follows a modified equal-weight S&P industry index [1]–[4], [8].

That difference in construction can matter more than the fee gap. The correct question is therefore not “which ETF is best?” but “which benchmark design best matches the intended semiconductor tilt, at an acceptable total implementation cost?”

Most importantly, a semiconductor ETF should be sized from the portfolio level. SPYM and QQQM already contain semiconductor exposure. The satellite adds to that look-through weight and can increase dependence on the same mega-cap growth drivers. Part 4 turns to the other major satellite in the wider implementation: the option wheel, where the central risk is not index overlap but the obligation created by selling options.

---

Appendix A: Semiconductor ETF calculation toolkit

A.1 Weighted portfolio fee

For ETF weights \(w_i\) and expense ratios \(e_i\):

A 10% SOXQ sleeve contributes approximately \(0.10\times0.19\%=0.019\%\) to total portfolio expense. A 10% XSD sleeve contributes approximately 0.035%. The 0.016 percentage-point portfolio difference equals $16 per year per $100,000 of total portfolio value before compounding.

A.2 Look-through exposure

Use current issuer holdings data because individual constituent weights change.

A.3 Herfindahl-Hirschman concentration index

The measure is most useful for comparing concentration consistently through time, not for declaring one weighting scheme universally safer.

A.4 Stress contribution

For multiple sleeves, sum deterministic impacts for a simple stress table. More advanced scenarios can model correlations and nonlinear option payoffs separately.

Appendix B: Source-verification notes

ETF facts come from issuer pages; methodology from Nasdaq; industry context from SIA/WSTS and peer-reviewed research. Refresh values before publication because fees, holdings and corporate actions change.

Appendix C: Workbook workflow for live holdings

Refresh issuer holdings, standardize identifiers, aggregate duplicates, calculate look-through weights and HHI, compare results with policy limits, and archive dated source files.

References

[1] Invesco, “Invesco PHLX Semiconductor ETF (SOXQ),” 2026. [Online]. Available: https://www.invesco.com/us/en/financial-products/etfs/invesco-phlx-semiconductor-etf.html. [Accessed: Sep. 11, 2026].

[2] BlackRock, “iShares Semiconductor ETF (SOXX),” 2026. [Online]. Available: https://www.ishares.com/us/products/239705/ishares-semiconductor-etf. [Accessed: Sep. 11, 2026].

[3] VanEck, “VanEck Semiconductor ETF (SMH),” 2026. [Online]. Available: https://www.vaneck.com/us/en/investments/semiconductor-etf-smh/literature/. [Accessed: Sep. 11, 2026].

[4] State Street Investment Management, “XSD: State Street SPDR S&P Semiconductor ETF,” 2026. [Online]. Available: https://www.ssga.com/us/en/intermediary/etfs/state-street-spdr-sp-semiconductor-etf-xsd. [Accessed: Sep. 11, 2026].

[5] H. Tan and J. A. Mathews, “Cyclical industrial dynamics: The case of the global semiconductor industry,” Technological Forecasting and Social Change, vol. 77, no. 2, pp. 344–353, Feb. 2010, doi: 10.1016/j.techfore.2009.08.003.

[6] Semiconductor Industry Association, “Market Data,” 2026. [Online]. Available: https://www.semiconductors.org/policies/tax/market-data/. [Accessed: Sep. 11, 2026].

[7] World Semiconductor Trade Statistics, “Recent News Release,” 2026. [Online]. Available: https://www.wsts.org/76/Recent-News-Release. [Accessed: Sep. 11, 2026].

[8] Nasdaq Global Indexes, “PHLX Semiconductor Sector Index (SOX),” 2026. [Online]. Available: https://indexes.nasdaq.com/index/Overview/SOX. [Accessed: Sep. 11, 2026].

[9] U.S. Securities and Exchange Commission, Office of Investor Education and Advocacy, “Updated Investor Bulletin: Exchange-Traded Funds (ETFs),” Feb. 23, 2023. [Online]. Available: https://www.investor.gov/introduction-investing/general-resources/news-alerts/alerts-bulletins/investor-bulletins-24. [Accessed: Sep. 11, 2026].

[10] Schwab Asset Management, "Schwab U.S. Dividend Equity ETF (SCHD)," 2026. [Online]. Available: https://www.schwabassetmanagement.com/products/schd. [Accessed: Sep. 15, 2026].

Technical companion

A thematic sleeve should be measured against the core by holdings overlap, weight, liquidity, expense and correlation. The concentration appendix retains the full HHI and effective-holding-count derivation used for that review.