The brief: Anthropic's S-1 has shown profitable, contracted demand for AI compute; SMH is at a technical bottom where the signals have started turning from red to green; buy a call on SMH today for a reversal to about $590, the midpoint between the 52-week high ($671.83) and the 22 October low ($520.53). This page ties the three pieces together and picks the structure. The author's first idea, a 45-day $600 call, is priced alongside: it is the wrong strike for a $590 target, because at the target it has no intrinsic value.
Update, 30 October before the open: closed at $8.75, before the Amazon print. Thursday SMH opened at $558.93, held the $555 cluster to the dollar ($555.50 at the low), bounced to $565.87 and faded through the afternoon as Alphabet fell 12.6%; the call was sold at 14:26 New York at $8.75 with SMH at about $558.40, for +$73 after commissions (+9%) on the $800 premium, two days after entry. The page's rule was a close under $555; going early was the right reading of the evening's risk: Amazon then printed AWS +39% with $75.9 billion of quarterly capex and fell 7%, Intel guided light and fell 26%, and SMH is indicated at $536 (−4.6%), where the call is worth about $5.30–7.30 (at 40–45% implied volatility): the exit kept roughly $150–340 per contract that the stop would have given back this morning. What the trade got right: the entry (the third up day, the scorecard turning, $549), the first two days (+44% at Wednesday's close, the $12 half-off level touched), and the discipline at the level. What it got wrong: the catalyst. The thesis said the Anthropic S-1 had turned AI demand into contracted revenue, and the hyperscalers confirmed exactly that (Alphabet Cloud +103%, AWS +39% with growth above 40% guided for 2027 and 2028); the market priced the cost of delivering it instead, and the chips, which are that cost, were sold with the spenders. A new long needs a close back above $538 and then the $562.8 flip before it is worth paying for; the set-up that would reopen one is the scorecard's, not the calendar's, and the next dated catalysts are AMD on 3 November and Nvidia on 17 November. The trade is recorded in the trade log.
Update, 29 October before the open: the first real test, at the $555 put cluster. Wednesday was the fourth up day: the Fed held (10–2, two dissents for a hike; a December hike is priced at 85%, the ten-year 5.03%), SMH closed $567.67 (+1.7%) with a high of $571.00, and the call closed at $11.50 (+44%), touching the $12 half-off level in the afternoon. Then the evening split the AI trade in two: Microsoft beat with Azure +48% and capex of $45 billion, below its guide, and rose 4%; Alphabet missed, posted −$9.1 billion of free cash flow, kept 2026 capex at $195–205 billion and guided 2027 to a "significant increase" (above $300 billion), and fell 10%. The market read Alphabet as the S-1's bear case written by a hyperscaler, costs scaling with revenue, and the chips followed pre-market: Marvell −10% (Lumentum deal talks), Micron and Intel −4%, AMD −3%, Broadcom −3%, Nasdaq futures −1.2%; SMH is indicated at $556.40 (−2.0%), on the $555 cluster. Wednesday's open interest made that strike the largest in the book (187 million dollars of put gamma, net −158), with the flip at $564.7 above it and the call strikes at $560–625 unchanged: above $555 the book dampens, below it dealers sell. At $556 the call is worth about $8.00, back to cost; the double from here needs $581 today, $586 by 3 November, $594 by 10 November. The decision is the close. A close under $555 is the stop (sell the call, about $7–7.50); a hold of $555 into Amazon and Meta tonight (AWS consensus +35–40%, the $496 billion backlog, the $220 billion capex year) keeps the trade alive, and Amazon's number is the one that can put the chips back on the front foot before AMD on 3 November and Nvidia on 17 November. If half was sold at $12 on Wednesday, the remainder is a free option on tonight.
Update, 28 October 09:50 New York, with Tuesday's open interest. SMH opened at $564.10 (+1.1%) on the FOMC day, exactly on the dealers' flip ($564.3), and the book is positive for the first time since 12 October (+$21 million per 1% in total, +$83 million in the next 35 days): the $555 put cluster is unchanged below the price and calls were added on Tuesday at $570, $580, $600 and the $625 wall. The scorecard reads seven of ten on Tuesday's close: confirmed. The 27 Nov $600 call is marked about $10.10 (+$210, +26%); the double still needs about $580 today, $587 by 4 November. What positive gamma changes for this position: above $555 the hedging now dampens moves and pins the stock near the big call strikes ($570, $580, $600), so the move to $590 has to come from the news (FOMC at 2 p.m., Microsoft and Alphabet tonight, Amazon and Meta on Thursday) rather than from the hedging flow that carried the first three days. The rules stand: half off at $12 now that SMH holds the flip, the rest at $16 on a $590 touch; stop on a close under $555 (raised from $530, since the floor the dealers defend has moved up with the open interest); out by 20 November unless above $600.
Update, 28 October before the open: the position is on, and the first day went its way. SMH closed Tuesday at $557.98 (+3.0%), every holding up, the third up day; the scorecard is six of ten and the dealer book is nearly flat (−$50 million per 1%, flip $561.7, the $555 cluster now below the price, net call gamma from $560 to the $625 wall). The author bought the SMH 27 Nov $600 call at $8.00 at 09:45 on Tuesday, the strike this page argued against for a $590 target, in the weekly that expires a week after the Anthropic debut; it is marked about $8.70 and has its own page under open trades. What that strike needs, at today's volatility: SMH at about $580 today, $587 by 4 November, $595 by 11 November, $604 by 18 November, $616 at expiry to double; the model gives the double a coin-flip chance (51%, median six days) with the stock 1% higher overnight. The rules for it: sell at $16 on the first touch of $590, or half at $12 once SMH holds above the $562 flip; stop on a close under $530; out by 20 November unless SMH is above $600. Today's FOMC (2 p.m.) and Microsoft and Alphabet (after the close) are the first test; Amazon and Meta on Thursday the second.
Pre-open update, 27 October 14:30 CET, with Monday's final open interest. The final figures revised the preliminary chain by +6.5% (1,438 contracts changed): the additions are tail protection far from the price (30 Oct $380 and $420 puts, 6 Nov $500 puts), the $530 wall thinned from 48,400 to 43,100 puts, and call open interest grew at $550–560 (about 5,000 contracts of November and January calls). Net dealer gamma is now −$212 million per 1% (−$339 million on the preliminary), the flip $562.5, and the largest single put strike is now $555 (ahead of $530 and $500), with net call gamma from $560 up. SMH is indicated at $548.80 (+1.3%) pre-market, inside the $550–555 cluster: the hedging flow turns in the position's favour a few dollars higher, not at the open. What the gap does to the structures (estimated from Monday's implied volatilities): the Dec 18 $550/$590 spread about $16.80 (break-even $566.80, 2.4× at $590 but only 1.4× on the maximum), the Nov 20 $560/$590 about $10.20 (break-even $570.20, 2.0×), the Dec 18 $560/$600 about $14.50 (break-even $574.50, 2.1× at $590, 2.8× at $600), the Dec 18 $560/$625 about $19.50 (2.3×). The recommendation stands with one change to the entry: do not buy the gap in the first minutes; buy if SMH holds above $545 (the first call gamma) through the first 30 minutes, and if it has already run through $555 by then, use the $560/$600 instead of the $550/$590, because the long strike should sit on the cluster, not above it. Stop and add levels unchanged ($530 close; $555 close). The scorecard is unchanged at 5 of 10.
The S-1 shows $16.3 billion of first-half revenue, a $974 million operating profit in the June quarter, positive operating cash flow, and take-or-pay compute contracts into the 2030s for Google, Amazon, Microsoft, Broadcom and AMD (floors of $417–637 billion). Every chip in it is made at TSMC and every accelerator carries memory. That is a demand confirmation for roughly a third of SMH by weight. It is also a slow catalyst with a dated end: after the 19–20 November debut the IPO becomes stock supply.
SMH fell 22% from the 6 October high, broke its $536–540 shelf on 22 October on capitulation volume, and on Monday closed back above it at $541.62 (+2.7%) with 96% of holdings up and the laggards leading. The reversal scorecard went from 1 to 5 of 10 in one session: shelf reclaimed, oscillator at zero and rising, participation, no new 52-week lows, Lam and Applied holding their lows. Still red: the candle, Nvidia and TSMC under their 50-days, memory flat on the week, dealer gamma.
With Monday's final open interest: put wall $530 (Monday's low to the dollar), the largest put strike now at $555 with $550 beside it (last week's protection), net call gamma from $560 up, the flip at $562.5, the call wall at $625 (58,600 November calls). Below $555 dealers sell rallies; above it they buy them. The long strike belongs on that cluster, the target under the call wall.
| Level | Price | vs close | What it is |
|---|---|---|---|
| Call wall | $625 | +15.4% | 58,600 November calls; the ceiling of the move, not the target |
| 20-day average | $598 | +10.4% | where the October slide started |
| Target | $590 | +8.9% | midpoint of the 52-week high and the 22 Oct low ($596); the 50-day is $578 |
| Gamma flip | $564 | +4.1% | above it the whole book is long gamma and pins; below it moves extend |
| Put cluster / first call gamma | $550–555 | +1.5–2.5% | the add trigger: a close above it is where hedging turns from selling rallies to buying them |
| Shelf (reclaimed Monday) | $538 | −0.7% | the $536–540 lows of 15–20 Oct; the failed breakdown needs it to hold |
| Put wall / stop | $530 | −2.1% | Monday's low; a close below it says the low is not in |
| 22 Oct low | $520.53 | −3.9% | the capitulation low; below it the next strikes are $500–505 |
| Structure | Cost | Break-even | Max gain | $545 | $560 | $575 | $590 | $600 | $625 |
|---|---|---|---|---|---|---|---|---|---|
| Dec 18 $550/$590 call spread the trade: capped at the target, long strike on the dealer put cluster | $15.43 $1,543 | $565.43 +3.8% | $24.57 (1.6×) | -15.43 -1.80 | -5.43 +2.47 | +9.57 +6.78 | +24.57 +10.84 | +24.57 +13.28 | +24.57 +18.17 |
| Dec 18 $560/$625 call spread for the overshoot to the call wall | $17.70 $1,770 | $577.70 +6.0% | $47.30 (2.7×) | -17.70 -4.30 | -17.70 +1.27 | -2.70 +7.57 | +12.30 +14.18 | +22.30 +18.55 | +47.30 +28.51 |
| Nov 20 $560/$590 call spread the catalyst window only; expires on the ANT debut | $8.87 $887 | $568.87 +4.4% | $21.13 (2.4×) | -8.87 -8.87 | -8.87 -8.87 | +6.13 +6.13 | +21.13 +21.13 | +21.13 +21.13 | +21.13 +21.13 |
| Dec 18 $560/$600 call spread the compromise: wider than the target | $13.18 $1,318 | $573.18 +5.2% | $26.82 (2.0×) | -13.18 -2.38 | -13.18 +1.63 | +1.82 +5.91 | +16.82 +10.15 | +26.82 +12.81 | +26.82 +18.43 |
| Dec 18 $570 call naked call: break-even is the target | $20.23 $2,023 | $590.23 +8.3% | uncapped | -20.23 -8.60 | -20.23 -2.55 | -15.23 +5.15 | -0.23 +14.45 | +9.77 +21.46 | +34.77 +41.33 |
| Dec 18 $600 call (the brief) the brief: strike above the target, pays only if $590 comes early | $10.82 $1,082 | $610.82 +12.1% | uncapped | -10.82 -6.50 | -10.82 -3.37 | -10.82 +1.14 | -10.82 +7.22 | -10.82 +12.17 | +14.18 +27.62 |
Mid prices of Monday's close (CBOE delayed chain); SMH implied volatility 36–39%, about the 40th percentile of the year, so premium is fair rather than cheap. The 20 November values use Black-Scholes with each leg's own implied volatility less two points for the fall in fear that a rally brings. Spreads are shown at their mid; expect to pay $0.20–0.40 more.
Solid: at expiry. Dashed: on 20 November with time left (the Nov 20 spread has none). The vertical lines are the stop ($530), the add ($555), the target ($590) and the call wall ($625).
The catalyst calendar cuts both ways. FOMC and Alphabet and Microsoft on Wednesday, Amazon and Meta on Thursday, AMD around 4 November, Nvidia on 17 November, the Anthropic pricing week of 16 November: the capex guides are what the thesis needs, and a soft one (or a Fed hike, priced at under 15%) would send SMH back to the put wall inside two sessions, because the book is still short gamma below $555. The scorecard is forming, not confirmed. Five of ten; the two that matter most are still red: Nvidia and TSMC under their 50-day averages, and the equipment names 17–24% under theirs after Lam's supply warning. A bounce that the leaders do not join fails at $555–564. The S-1 is a slow catalyst with a dated end. It confirmed demand for the names it contracts; it did not change the Lam supply story or the AI-financing story, and after 20 November a $100 billion primary competes for the same dollars. Implied volatility is fair, not cheap (36–39%), which is why the structures are spreads: a naked call pays for a volatility that would fall on the rally it needs.
Data: Monday 26 Oct 2026 close (Yahoo), CBOE delayed chain with Monday's final open interest (preliminary ~07:00 CET, final ~14:15 CET on 27 Oct; the final revised the preliminary by +6.5%), the SMH health page and the Anthropic S-1 research. Quotes must be re-checked at the open; a 2% gap changes every number above.