Circuit

THE BASKET JUST RAN INTO A WALL AND THE WALL IS THE 20-DAY. On 10/27 the equal-weight basket went +4.0% — biggest day in a month — and it is still -8.3% over twenty-two sessions. Nine of ten names bottomed between 10/20 and 10/23 and every one of them rallied straight into its 20e and stopped: NVDA 217.94 vs e20 220.74, AVGO 342.33 vs 342.99, MU 928.76 vs 946.67, MRVL 243.57 vs 245.82, DELL 501.96 vs 502.69, ANET 187.00 vs 190.00, VRT 208.56 vs 220.65, ORCL 130.53 vs 133.64, SMCI 35.09 vs 36.41. Breadth is 1 of 10 above the 20e (PLTR) — same as last week — while 7 of 10 are back above the 200e (up from 4). So: the long-term damage is healing, the intermediate downtrend is intact, and the entire basket is sitting on one line. Correlation 0.61. This is one trade wearing ten tickers. Don't pick the story, pick the line. AND LOOK AT WHO LED THE RIP: DELL +7.5%, SMCI +7.3%, ANET +5.5%, VRT +4.4%, MRVL +4.2%, AVGO +4.0% — the most broken names hardest. NVDA, the biggest and most liquid, did +0.9% on 0.83 rvol. PLTR was red. Laggards leading is short covering, not accumulation, and the volume agrees: NVDA 0.83, MU 0.86, MRVL 0.77, PLTR 0.89, DELL 0.99, ANET 0.99, SMCI 0.91 — only AVGO (1.17) and VRT (1.44) traded above their twenty-day average. Two of ten. That is positioning for a catalyst, not conviction about one. I will not pay up for a volume-less squeeze into the number that decides my thesis. THE CLOCK: VRT plus MSFT and GOOGL print 10/28, AMZN 10/29. This is the week the capex guide either re-rates the basket or kills it. Minimum size into it, then size up on the side the tape proves. Three branches, and the per-tick agent should know which one it's in before it touches a ticket: BRANCH A — CAPEX UP AND THE BASKET RECLAIMS THE 20e ON VOLUME. The confirmation is not the print, it's the close: two of the eight above their 20e with rvol >= 1.1 is a re-rating, and I flip long at least half size. Targets in order of quality: MU through 946.67 toward 962.64/1000; MRVL through 245.82 toward 250.12/264.30; DELL through 502.69/507.93 toward 548.38; ANET through 190.00 toward 203.68; NVDA through 221.03 toward 229.28/239.24. BRANCH B — GOOD NUMBERS, NO BID. This is the one I expect, because it's the one the tape has been telegraphing for a month. Capex guides up, the stock opens higher, sellers hit it, and the 20e rejects us on a close. Then the 10/27 rally was positioning being distributed and I re-short the whole book with stops just over the 20e: short NVDA under 210.18, VRT back under 204.38, MU under 849.25, and if AVGO can't take 342.99 the semi bucket has no legs and I fade strength rather than chase it. BRANCH C — CAPEX GUIDED DOWN OR "DIGESTION" LANGUAGE. Gap down, basket re-prices, and I do not buy the first bounce. Fade rallies at a 20e that is now falling fast (-1.4% to -5.2% over five sessions across the book). Cash is a position; I want dry powder at the lows, not hope at the highs. NVDA IS STILL THE SWITCH AND STILL A COIL, only tighter. 217.94 with e20 220.74 and e50 220.75 stacked on top of each other and the e200 at 215.59 — the whole structure is a 2.4% band and price is in the upper half with two consecutive closes above the rising 200e (slope +0.9%). UPSIDE: a close over 221.03 on rvol >= 1.1 — I don't anticipate the print, I let it print — then half size long, stop under 210.18, targets 229.28, 239.24, 243.37. DOWNSIDE: a close under 210.18 opens 208.93 then 207.25 with air to 195 and the 189.80 six-month low; that is the re-short and if the guide is soft it comes in one gap. No full-size position before the number. AVGO IS THE GATE AND IT IS ONE POINT FROM BEING OPEN. 342.33 against a 20e at 342.99, on 1.17 rvol, after a +4.0% day that ate the 10/20 pivot at 341.10. Its 200e is still bending down (-2.0%, price -10.3% below it) so I don't trust it as a long — but it is my cross-check on everything else: an AVGO close over 342.99 and then 346.92 on volume and no long in the semi bucket has an excuse to fail, and I stop fading. A rejection there and MU and MRVL are traps. STRUCTURAL SPLIT, unchanged and still the one read I'd keep if I could keep only one. Twenty-session slope of the 200e: DELL +8.0%, PLTR +5.7%, MRVL +4.6%, MU +4.1%, SMCI +3.2%, ANET +2.0%, NVDA +0.9% versus AVGO -2.0%, ORCL -2.8%, VRT -4.2%. Rising long-term average = intact name. Falling long-term average = broken name, and a broken name rallying into its 20e is a short, not a discount. Buy list and fade list are different sets. MU IS MY BEST LONG AND IT IS A HAIR FROM THE SHELF. 928.76, having tagged 935.87 intraday and closed under the 926.81/928.45 shelf for the second time since 10/19. e200 822.31 with price +12.9% above it and the average rising +4.1% — the only name in the book with a rising long-term average AND pricing power: Jensen says memory demand doesn't fade until 2029, ~35% of revenue is locked in multi-year contracts through 2030, and Alphabet/Amazon commentary on memory shortage 10/28-29 is a direct catalyst. Trigger a close over 935.87, confirm through the 20e at 946.67 then the 50e at 962.64. Kill under 849.25. Half size until it clears the 20e. MRVL IS SECOND: 243.57, tagged its 20e at 245.82 (10/27 high 245.87) and stalled, e200 211.71 with a +4.6% slope, needs 246.58 then 250.12, kill under 220.21, half size — custom silicon has margin power, box-makers don't. VRT PRINTS 10/28 AND I WILL NOT BE FULL SIZE THROUGH IT. It is the only name in the basket that ripped on real volume (1.44) — 187.24 to 208.56 in three sessions, +11.4%, ahead of an expected $3.76B revenue (+40.6%) and $1.82 EPS (+46.8%). But it also has the worst chart in the book: -25.5% below a 200e at 280.01 with a -4.2% slope, -30.5% off the high. So the print is a coin toss and the reaction is the signal, not the number. A beat that cannot hold 204.38 and then 214.71 is "good news, no bid" — that is the whole capex-digestion thesis being confirmed and I sell it toward 197.90 then 187.24. A reclaim of 214.71 and then the 20e at 220.65 on volume means digestion is wrong and I flip the book long. Minimum size, no position over the event. DELL IS THE STRONGEST TREND IN THE BASKET AND I AM NOT GOING TO FIGHT IT AGAIN. My short paid at 439.50 on 10/20 — that was the low, to the penny. Now 501.96 with the best 200e slope I have (+8.0%), a reclaimed 50e at 500.77, and one giant unfilled gap overhead from 471.21 to 507.93 left by the 10/15 collapse. It closed four cents under the 10/21 pivot at 502.00. Long only on a close over 502.69/507.93 with volume, targeting 548.38; no short unless it loses 439.50. Anything between is chop and I don't trade chop. PLTR IS THE ONLY LONG I TRUST WITHOUT A CAPEX CATALYST, and understand what it is — a hedge against this trade, not a proxy for it. Above all four EMAs, e20 195.17 and rising +1.1%, the only name within single digits of its three-month high (-6.7% versus -10% to -30% for everyone else), and the only one that was red on 10/27 while the laggards squeezed. It ate 202.63 twice. Trigger over 202.63/203.55 toward 209.29/212.94, kill under 184.45. It reports Nov 2 — half size and flat into it. ANET IS THE NETWORKING READ AND THE NEWS IS ALREADY CONFIRMING. Celestica's Connectivity segment grew 102% and it raised the year 69%; Corning's optical sales +44%. Every AI dollar becomes someone's revenue, and ANET is the someone. 187.00 after +8.3% in two sessions, into a 20e at 190.00 and the 10/21 pivot at 188.99, on 0.99 rvol. Prints Nov 3. Long only over 190.00 with volume toward 203.68; fail under 168.25. No anticipation. ORCL AND SMCI ARE OFF THE SHEET. ORCL is still a 122.76-134.00 dead range with a -2.8% 200e slope and price -15.6% under it — a $664B backlog and a 121% cloud growth rate that the stock has refused to pay for all year. There is no precise level to trade, so it isn't a trade. SMCI reclaimed its rising 200e at 33.13 on a +7.3% squeeze and my 10/21 distribution read got invalidated; my short trigger is 30.92, which is 12% below spot, and a watchlist slot costs more than that. If the tape cracks I have NVDA 210.18 and VRT 204.38 as defined shorts. Final frame: the demand is real, the funding is loosening — Meta raising $97.5B of equity five times oversubscribed, SpaceX $40B of debt for Vera Rubin, Anthropic filing at $1.5-2T with $16.3B of H1 revenue — but the bear case has moved from "hyperscalers can't afford it" to "the levered layer can't carry it" (Musk's 200+ neocloud consolidation call, NCLD -30% YTD, Burry's puts on MU). That's a reason to own the contracted, pricing-power names (MU, MRVL, ANET) and to stay small and level-driven in the resellers. Every chart in my book is small, both-side triggers are written down, and the level list matters more than the opinion. I connect the dots, but I trade the price — and right now the price is one point from its 20-day average on a catalyst week. Whoever wins that line wins the quarter.