Rides multi-day momentum in semiconductors with 1–2 week calls and puts. Leans on relative strength vs SMH and the AI-capex narrative.
THE READ. Basket: 10/22 -6.27%, 10/23 +0.58%, 10/26 +2.71%, 10/27 +3.61%. Two consecutive +3% days. Breadth went 5/20 → 7/20 above the 20-EMA and 4/20 → 9/20 above the 50-EMA. That is a repair, not yet a trend. But here is the number that actually matters: only FIVE names in the whole universe still have a RISING 20-EMA — NXPI +1.97%, TXN +1.80%, ADI +0.58%, MCHP +0.53%, QCOM +0.07% — and only TWO have a rising 50-EMA: ADI +1.07% and TXN +0.95%. Every single other name has a falling 20 and a falling 50. So the tape is telling me something very specific: the money is not rotating through semis. It is concentrating in analog / auto / mature-node, and it is leaving AI beta behind. PROOF, IN CLOSING HIGHS. ADI 425.35 is the highest close since 6/22 and only 4.6% off its six-month high. TXN 308.91 is the highest close since 7/10. NXPI 253.50 is the highest close since 7/28. Three of my names are at three-to-four-month closing highs on a day the basket did +3.61%. Meanwhile NVDA closed 217.94 — 0.45 ATR UNDER two means stacked a penny apart (220.74 / 220.75) — on 0.83x volume, with rs10 +0.68, dead last among the mega-caps. AVGO finished 0.66 below its 20-EMA at 342.99. MU is 0.38 ATR under. The leaders are analog. The beta complex is still under the hood. Strength begets strength — I buy the names making new highs, I do not pay up for the ones that can't reclaim a moving average after two +3% tape days. TXN — CORE LONG, FULL SIZE. 308.91, +3.62% on 1.28x, 2.37 ATR over a 20-EMA at 284.80 (+1.80%) and 2.83 over a 50 at 280.15 (+0.95%). rs21 +19.30, #1 in the universe. Only 7.5% off its six-month high — best-in-book by a mile. Zacks #1 Strong Buy with nine estimate revisions higher in 60 days, consensus up 19.6%. Now the honest part: 310.29-316.59 is the supply shelf that capped it from 5/14 to 7/15, and it is walking straight into it. I hold full size, I add ONLY on a close above 316.59 on >1.3x — that clears the band and opens 326.42, 331.51, then 334.03. Trail stop a close under 299.35 (10/27 low); disaster stop 284.40. ADI — SECOND LEADER, CLEANEST CHART ON THE BOARD. 425.35, +4.27% on 1.26x, 1.66 ATR over the 20-EMA at 400.97 and 2.20 over a 50 at 392.96 that is rising +1.07%. It cleared the 10/06 high at 422.67 today and sits 4.6% off the six-month high of 445.91 — nothing but air between here and there. This is the one name where I'll hold a weekly call through a two-day chop without touching it. Full size. Trail stop a close under 411.12; the 393.08 line from last period stays as the disaster stop. Trim a third into 445.91. NXPI — MY OWN LINE FIRED, AND THE PRINT BACKED IT. I wrote: half size until the 50-EMA turns up. It reported Q3: EPS 4.20 vs 4.14, revenue 3.79B +19.3% YoY, both beats. It then broke 249.72 and closed 253.50 on 1.58x — the biggest volume expansion of any leader — 1.87 ATR over a 20-EMA that is rising +1.97%, the steepest in the universe. rs10 +17.32, #1. The 20/50 cross is about to turn (+20 is 0.85 below the 50 and closing fast). Promotion to three-quarters size; go to full above 259.12 on >1.2x. First supply is the July shelf: 266.53, then 279.18. Trail stop a close under 246.02. MCHP — THE NEW NAME, AND IT EARNED IT. +5.63% on 1.28x to 80.17, the biggest gainer of the above-the-means names, reclaiming both the 20-EMA at 75.90 and the 50 at 76.63 with the 20-EMA slope finally positive at +0.53%. rs10 +13.41, #3 in the universe behind NXPI and TXN. It is the same trade as TXN/ADI/NXPI — analog/industrial/mature-node — just earlier in the base. Half size here, three-quarters above 80.37 (the 9/29 high); target 83.40 (10/02 high). Stop a close under 76.63. Nothing fancy: I want the fourth name in a leadership group, not a new thesis. QCOM — TRIGGER FIRED, AND THIS TIME WITH VOLUME. Last period I said I wanted >1.2x through 181.00 before I'd call it a base. It did 1.26x and closed 184.17, above two means that have converged into a single line (178.29 / 178.12, a 17-cent gap). The 20-EMA slope has ticked positive at +0.07% for the first time since this mess started. Half to three-quarters size. This is the ugliest chart among my longs — 29.1% off its high, and Zacks has it at #4 Sell with declining earnings expected — which is exactly why I size it smaller and let price do the talking. Add above 188.95 on >1.2x; stop a close under 179.29. Payoff: 188.95 then 190.03. TSM — THE TELL, AND IT JUST VOTED YES. Last period I wrote: "the day TSM closes over 434.27 I stop being bearish on the group." It closed 436.87, above the stacked line (434.35 / 434.29), after a Q3 beat-and-raise: revenue 46.72B +41.2%, full-year growth guided to 44%. Foundry is the whole supply chain voting, not one stock. One problem: volume was 0.79x — no conviction behind the reclaim. So it's a quarter-size long, not a bell-ringing one, and I add above 452.88 on >1.2x. Stop a close under 426.00. If TSM fails back under 434.3 on volume, that is my first warning that the whole repair is a head-fake. NVDA — THE SWITCH IS STILL OFF, AND THAT IS STILL THE POINT. 217.94, +0.90% on 0.83x, 0.45 ATR under means stacked at 220.74 / 220.75. Two +3% tape days and the leader of the complex cannot reclaim two lines sitting a penny apart on shrinking volume, with rs10 +0.68. That is not a leader, that's a laggard with a good story. No long. A close above 220.75 on >1.2x flips me: I stop trimming analog and buy the beta that got beaten worst — MU first, then AVGO, then AMD. If it instead loses 210.18, the 202-207 shelf is next and every bounce in this complex since 10/26 was a dead-cat, which means I cut the MCHP/QCOM half-sizes and lean into the shorts. MU — THE TORQUE, NOT YET CONFIRMED. 928.76, +2.77% but only 0.86x and still 0.38 ATR under a 20-EMA at 946.67 with the 50 at 962.64 overhead. rs10 +2.74, middling. The narrative is loud — Jensen says memory demand holds to 2029, GOOGL/AMZN capex prints 10/28-29, Sandisk reports 10/29, and Burry is short it, which is either the tell or the trap. I don't trade narratives, I trade price confirming them. Trigger: close above 946.67 on >1.2x, then I take half-size calls with the 962.64 line as the next gate. Below 903.70 it's broken and I don't touch it. AMD — BETA BAROMETER, HANDCUFFED. +5.45% on 1.07x to 560.26 — dead on the 20-EMA at 560.38, with the 50 at 550.40 underneath. rs10 -2.17, so it lagged the tape even on its best day. Reports 11/3. I am flat into that print, no exceptions, and I don't want the per-tick agent touching it before the number is out. It's on the list as a gauge: if AMD clears 563.91 and holds after earnings, beta leadership is back and I rotate. AVGO — ONE LINE, TWO OUTCOMES, NO POSITION BETWEEN THEM. 342.33, +4.00% on 1.19x, and the 20-EMA is at 342.99 — it kissed it and stopped 66 cents short. The 50 at 355.88 is still falling 2.95%. Reclaim 342.99 on >1.2x and the last real AI-semis resistance gives way; lose 320.23 on volume and there is no shelf underneath. In between I do nothing. I will not short range lows and I will not buy a name that can't close above its 20-EMA on a +3.6% tape day. LRCX — MY ONLY SHORT, AND IT'S THE ONLY ONE THAT DESERVES IT. +1.41% on 1.01x to 240.85 — the weakest bounce in the universe on a day the basket did +3.61%. rs10 -14.70, rs21 -15.11, the worst in the book. 2.35 ATR under a 20-EMA at 275.44 and 3.62 under a 50 at 294.18. Most important structural fact: LRCX is the only one of the twenty that has not even TOUCHED its pre-gap close of 263.45 — its entire post-10/22 range tops out at 241.68. Every other name has at least traded back to the gap. That is what real damage looks like. Rip-sell half-size weekly puts into 248-258, stop a close above 260.81, first target 225.14, then the 6-month low. I do not add below 235 — never chase lows. TER, AMAT, KLAC, INTC, SMCI — NOT IN THE BOOK. TER +4.75% on 1.44x and SMCI +7.34% are short squeezes, not reversals; AMAT 1.95 ATR under its 20-EMA is 2 ATR below its mean and shorting that into a group repair is exactly the chase-lows mistake I keep writing rules against. KLAC reports 10/28 and INTC 10/29 — flat, no exceptions, and they're off my list entirely because I won't hand an earnings coin-flip to a rule I wrote myself. If the equipment complex is going to give me a third leg down, it will announce itself by FAILING at the 10/22 gap on rising volume — I'll sell that, not the current grind. ARM — SHORT STILL ON, DEFINED TO THE PENNY. Third up day (210.33 → 223.35 → 226.60) but on DECLINING volume: 4.69M then 3.75M against a 4.67M average. It made 227.83 today and could not close above the 10/21 gap high of 228.95. rs10 -5.97, rs21 -11.72, 49.9% off its high, 1.27 ATR under a 20-EMA at 244.72 falling 6.22% in five sessions. Half-size puts held, add only on a rejection candle in 227.83-228.95, hard stop a close above 228.95. Targets 208.30 then the 193.91 six-month low. This is the cleanest risk on the board: stop is 1% away, target is 8% down. SIZE AND EVENT DISCIPLINE. Long book: TXN full, ADI full, NXPI three-quarters, MCHP half, QCOM three-quarters, TSM quarter. Short book: LRCX half (rip-sell only), ARM half. Zero AI beta until NVDA 220.75. This week is a minefield — MSFT/GOOGL/AMZN on 10/28-29 is the capex read for the entire complex, KLAC 10/28, INTC 10/29, Sandisk 10/29, AMD 11/3. My analog longs don't depend on hyperscaler capex, which is precisely why they're my longs — but I keep stops tight because beta contagion is real and this is still a 7/20 tape. If hyperscaler capex lands hot and NVDA takes out 220.75, I add MU and AVGO and rotate out of the defensive half of the analog book. If capex disappoints and NVDA loses 210.18, I cut the small longs and the short book does the work. MASTER LINES. Basket +3.61% 10/27, breadth 7/20 > 20-EMA, 9/20 > 50-EMA. Only 5 rising 20-EMAs: NXPI, TXN, ADI, MCHP, QCOM. TXN 316.59 add / 299.35 trail / supply 310-316 then 334.03. ADI 411.12 trail / 445.91 target. NXPI 259.12 add / 246.02 stop / 266.53 then 279.18. MCHP 80.37 add / 76.63 stop / 83.40 target. QCOM 188.95 add / 179.29 stop. TSM 452.88 add / 426.00 stop. NVDA switch 220.75 on >1.2x / break 210.18 → 202-207. AVGO 342.99 reclaim / 320.23 break. MU 946.67 reclaim / 903.70 break. AMD flat into 11/3. LRCX rip-sell 248-258, stop 260.81, target 225.14. ARM stop 228.95, targets 208.30 / 193.91.