Why a call on a semiconductor ETF is always cheaper than a basket of calls — and why that is not a Black–Scholes failure
Data as of 15.10.2026.
In short. A Substack post from 25.04.2026 argues that deep OTM January 2028 LEAP calls on the semiconductor ETF SMH are "underpriced" versus calls on individual chip names, because Black–Scholes "doesn't account for correlation". The claimed edge isn't there. A portfolio of options is always worth at least as much as an option on the portfolio (Merton, 1973), so an SMH call has to be cheaper under any model. And on average implied correlation sits above realised, which means index vol is rich to its components, not cheap. SMH calls are a legitimate convex sector bet, but not an arbitrage.
Why this matters for the post: when correlation is cheap, as in July, index options are the ones that look relatively cheap. The author wrote in April, before that low. Since then correlation has repriced by two thirds, so the "ETF vol is cheap" argument is even weaker in October than it was in April. COR3M is an S&P 500 index, not an SMH one, so it gives direction, not a reading for SMH.
What it is
A Substack post (by UnsolicitedNotFinancialAdvice, 25.04.2026) argues the following. Implied volatility (IV) on options on the semiconductor ETF SMH is below the weighted-average IV of its components. Therefore January 2028 LEAP calls on SMH, especially deep OTM, deliver "≈18% (ATM) and almost 20% (OTM) more alpha" than the same calls on single names. The author attributes this to Black–Scholes "not accounting for correlation".
What the object is: - Domain: US listed options, OCC clearing. - Class: long calls on a sector ETF. - Wrapper: a standard exchange-listed ETF option; the underlying is a basket of 25 chip companies, ≈66% of it in the top twelve, per the author.
Positioning
Niche: options on semiconductor sector ETFs. SMH leads: ≈$70.6bn in assets vs ≈$45.9bn for SOXX (etf.com, July 2026); on 14.10.2026, $68.98bn vs $42.65bn (etfcentral). SMH options trade over 800k contracts a day with over 5m in open interest (apexvol.com, 15.10.2026). SMH's share of sector-ETF options volume is not verified: that needs OCC or Cboe per-ticker data. Market: global — a US ETF holding the world's chipmakers (TSM, ASML-class); WSTS's 06.08.2026 forecast has the semiconductor market growing +108% in 2026, and US listed options volume is up +16% year-to-August 2026 versus 2025 (OCC). Market stage: the underlying sector is in scaling on the edge of a transition (confidence 0.55): semis are the most crowded trade in the market, the silicon layer is already handing over to the next stage, and neoclouds are in a Minsky exit phase; on the tape this shows up as the fourth 10%+ correction since June, a frequency not seen since 2002 (GuruFocus, 15.10.2026). The ETF options market itself is in mature scaling (confidence 0.7). Macro stage: macro regime supportive but fragile (reading of 19.07.2026, now 89 days old and close to stale); the fragility back then came precisely from implied correlation at a 16-year low (COR3M 7.2), and by 15.10.2026 it was 12.04 (+67% over three months, Yahoo Finance) — a gradual climb rather than a spike over a few sessions; phase: late cycle with fragility, credit still calm. Timing conclusion: unknown — the "edge" the author describes is an arithmetic identity, not mispricing, so there is no timing for it; timing a sector bet via calls is a vol question: SMH 30-day IV 35.7% with IV rank 37, i.e. mid-range (apexvol, 15.10.2026); next check 13.11.2026.
What we found
1. The core point: the "cheapness" of the ETF option is a theorem, not a Black–Scholes failure. - A portfolio of options is always worth at least as much as an option on the portfolio (Merton, "Theory of Rational Option Pricing", Bell Journal of Economics, 1973). So an SMH call must be cheaper than the weighted basket of component calls under any model. - Basket variance is σ²_ETF = ΣΣ wᵢwⱼρᵢⱼσᵢσⱼ. As long as correlation ρ < 1, ETF IV is below the weighted-average component IV, and Black–Scholes has nothing to do with it: the model is applied to each underlying with its own IV. - Illustration: the Herfindahl index of SMH weights is ≈0.07 (estimate: NVDA around 20%, TSM and AVGO roughly 8–10%, no separate source). Assuming average component IV of 50% (an assumption, not a reading), SMH IV of 35.7% implies a correlation of ≈0.47. - What the author flagged as "only TXN has lower IV than the ETF" is exactly this formula at work.
2. The real question is whether implied correlation is below future realised. The evidence says the opposite. - On the S&P 500, average implied correlation is 39.5% vs 32.5% realised; on the DJ30, 46.0% vs 35.5% (Driessen, Maenhout, Vilkov, Journal of Finance 64(3), 2009). Index options are on average rich to their components: that's the correlation risk premium, and broader readings put it in the same +7…+18 point range. - Buying ETF calls instead of the basket doesn't harvest a market error — it pays that premium. Dispersion funds do the reverse: sell index vol, buy single-name vol. - Caveat: the data is S&P 500 and DJ30, not SMH. No separate reading was found for a narrow sector where NVDA weighs around 20%, nor for deep OTM calls. That needs historical IV surfaces for SMH and its 25 components (OptionMetrics / Cboe DataShop, paid).
3. "Lambda = alpha" is the wrong yardstick. - Lambda (elasticity Ω = Δ·S/C) is higher on the ETF option precisely because it's cheaper at lower IV. - But the ETF underlying also moves less, so the option's own volatility ≈ Ω·σ comes out similar for the ETF and for a single name. - Alpha is expected return above the risk-fair return; elasticity is just leverage. "18% more alpha" means 18% more leverage on a less volatile underlying. - The post has no methodology: the author only says "I just took a default value from IBKR, for simplicity". Tables are posted as images, with no snapshot date, strikes or premiums. The author's numbers can't be verified and don't enter the calculation.
4. Deep OTM SMH LEAP calls as a sector bet — legitimate, but an ordinary convex bet. - Loss is capped at the premium, the payoff is convex. The case for it is a view on the sector, not an "edge". Against it: crowding in semis and a string of corrections. - SMH price: $556.63 on 31.08 → $609.00 on 30.09 → $558.84 on 15.10.2026, i.e. −7.4% over four sessions from $603.33 on 09.10 and −8.2% from 30.09 (stockanalysis.com; MarketChameleon shows the same price on 15.10). - The IV disagreement across sources is a finding in itself (Fig. 4). Anyone buying these calls should take IV from the actual order book, not an aggregator. - SMH calls just add the same semis bet through a second wrapper. For someone already holding chip stocks or sector ETFs, it's not a new bet.
Who it suits and when it breaks
- "SMH LEAP calls as an arbitrage against the components" — no: there is no edge, the thesis is wrong.
- SMH calls as a convex sector bet — yes, if it's long options only, without margin: the loss is capped at the premium.
- What breaks the structure: selling uncovered legs; poor liquidity in far January 2028 series (wide spread to mid); moving to a margin account.
What kills it, and at what number
- Scenario 1 — the post's thesis is wrong as an arbitrage (base case). On 13.11.2026, pull 30-day ATM IV for SMH and NVDA from apexvol on the same day. The diversification identity holds if SMH IV < NVDA IV. If SMH IV ≥ NVDA IV, the correlation argument fails for this basket and the analysis gets rewritten.
- Scenario 2 — correlation regime. On 13.11.2026, pull the COR3M close (Yahoo/Cboe). The reading "July's 7.2 low was an extreme, not the norm" holds if COR3M > 7.2. A return to ≤ 7.2 would mean cheap correlation is a persistent state. Then the author's cheap-ETF-vol argument regains a market footing, though as a "Black–Scholes failure" it is still wrong.
- Scenario 3 — a sector bet via calls. In the worst plausible case the whole premium goes to zero: −100% of what's put into the option. For deep OTM that's the base case, not the tail: breakeven needs a big move within 15 months. What kills the bet: SMH below the strike on 15.01.2028 — for "deep OTM" that's the majority outcome.
What we don't know
- Not verified: historical IV surfaces for SMH and its components (OptionMetrics or Cboe DataShop, paid). Only they would show whether implied correlation specifically in SMH, on deep OTM calls and at LEAP tenors, came in below realised. It's the only measurement that could rescue the author's thesis.
- Not verified: SMH price on 24–25.04.2026. Aggregators returned only August–October, so it's unknown how the author's idea has played out since April.
- Not verified: the author's numbers. The IV and lambda tables are images; strikes, premiums and snapshot date aren't given.
- Unknown: the source of the 55.71% IV with rank 90.47. The 16.10 search snapshot doesn't name it; Barchart and optionsamurai returned empty pages.
- Unknown (no instrument): an implied correlation index for semiconductors. Cboe only publishes COR1M and COR3M for the S&P 500; nobody publishes a sector one.
Bottom line
The post can't be dismissed outright — there is substance — but the claimed edge isn't there: a cheaper option on the basket is Merton's theorem, not a model failure, and on average index options are rich to their components, not cheap. What remains useful: - a long call on an ETF is the cheapest way to buy sector convexity, but it's cheaper because it carries less convexity; - "higher lambda" is not alpha; - after a big run in a sector, look at IV percentile, not just absolute IV.
Check calendar
| Date | Object | What to pull | Where | Forecast |
|---|---|---|---|---|
| 13.11.2026 | SMH vs NVDA, 30-day IV | 30-day ATM IV for SMH and NVDA on the same day | apexvol.com/options/smh and apexvol.com/options/nvda | SMH 30-day IV will be below NVDA 30-day IV (diversification identity; a miss refutes the post's correlation reading for this basket) |
| 13.11.2026 | COR3M, S&P 500 implied correlation | COR3M daily close | finance.yahoo.com ^COR3M or Cboe | COR3M closes above 7.2 (July's 16-year low was an extreme, not the norm; a miss means cheap correlation is persistent) |
Sources: Substack post 25.04.2026 · apexvol SMH 15.10.2026 · stockanalysis SMH history · MarketChameleon SMH · Yahoo ^COR3M · optionsamurai SMH IV · Driessen, Maenhout, Vilkov 2009, JF · etf.com SMH vs SOXX 2026 · etfcentral SOXX vs SMH · WSTS, 06.08.2026 · GuruFocus 15.10.2026
Research for information only. Not investment advice.