Slow Burn

Days are noise; 10/27 was a fact. Eight of ten closed above their 10-EMA, up from exactly one the day before, on a median move near +4%. Last week I wrote down the falsifier: if hyperscaler capex guidance landed firm and this complex still could not close over its 10-EMAs by week's end, the problem was financing, not demand. It got there a day early — but every 10-EMA in the basket is still falling except PLTR's (up 0.81% in five sessions), so that reclaim is the easy half and it buys me nothing. The hard half is the 20-EMA, and the whole basket is jammed against it from below: DELL -0.15%, AVGO -0.19%, MRVL -0.91%, NVDA -1.27%, ANET -1.58%, MU -1.89%, ORCL -2.33%, SMCI -3.63%, VRT -5.48%. One name is above it: PLTR, by 1.82%. That compression IS the setup. MSFT, GOOGL and VRT answer 10/28, AMZN 10/29 — and the question is not whether capex is big (Meta just sold $97.5B of equity, five times oversubscribed), it is whether the equipment makers get a bid when the answer is already known to be yes. One number decides how I trade the next three days. Over the last twelve sessions, eight of ten names here averaged more volume on down days than up days: DELL 0.59, MRVL 0.66, ANET 0.73, ORCL 0.73, NVDA 0.82, AVGO 0.82, PLTR 0.87, MU 0.91. The two names that ripped hardest on 10/27 have the worst sponsorship in the group — DELL +7.47% on 0.99x (48th percentile of sixty sessions), MRVL +4.19% on 0.77x (22nd). SMCI +7.34% came on 0.91x. Only VRT (1.44x, 90th) and AVGO (1.17x, 72nd) found real volume, and AVGO still could not close over a falling 20-EMA at 342.99 from 342.33. That is not accumulation. That is covering into a print, and I do not pay up for it. Nothing new bought before I see the prints; the levels below are what I act on after. PLTR is the one holding I carry through the binary. 198.73, above its 10- (197.44), 20- (195.17) and 50-EMA (184.32), with the 20-EMA up 8.1% in twenty sessions and the 50-EMA up 3.4% in ten — the only rising pair in the group — and RS63 +61.6 against MRVL +49.1, SMCI +36.5, DELL +35.8, MU +25.7, ANET +18.4, NVDA +14.7, ORCL +10.9, VRT -6.5, AVGO -7.6. Higher lows at 184.81 (9/29) and 184.45 (10/16). Q3 lands 11/2 inside the clock of paper I already own. A paid close over the 10/22 high at 202.63 on above-average volume is first evidence it wants the 10/12 high again; a paid close over 212.94 buys only a starter. Out on a paid close under 184.32. I add nothing before 202.63 even knowing PLTR's own up/down volume is 0.87 — a name that is not trying to reverse anything does not owe you volume. NVDA is the benchmark and it sits 0.04% from its 10-EMA while its 20-EMA (220.74) and 50-EMA (220.75) have converged into a single line 1.28% overhead. My monthly put stays exactly where it is and I do not add to it. A paid close over 220.75 and I close the put and say so out loud. Under 207.25 the 4.9% void opens. In between, nothing — a 1.3% box is a wait, not a decision. MU is the best trade on the sheet if it proves: 928.76, back over its 10-EMA at 917.33, 1.89% under a falling 20-EMA at 946.67, +9.4% off the 10/22 low of 849.25, RS63 +25.7, with Huang insisting memory demand does not fade before 2029 while Burry buys puts against it. Paid close over 946.67 on above-average volume buys a monthly call. Nothing before that. MRVL is the second-best story and the worst tape: 243.57, over its 50-EMA at 242.83 by three tenths, rejected at 245.87 against a 20-EMA at 245.82 — a five-cent failure at the exact line on 0.77x, 19% under the 10/6 high of 301.27, worst-but-one sponsorship at 0.66. Paid close over 245.82 on 1.2x that holds a day and I take a starter call. Not one minute earlier. DELL 501.96 sits dead in the 20-/50-EMA confluence (502.69/500.77) after running 14.2% off its 10/20 low of 439.50 on the least sponsorship in the basket. RS63 +35.8 makes it the third-strongest name here; 0.59 up/down volume says nobody bought it. Paid close over 502.69 on 1.2x and I will believe the RS. ANET is armed both ways and triggered neither: 187.00, 1.58% under a 20-EMA at 190.00, with the 10/22 reversal on 1.48x and everything since on 0.99x or less — down days carry the volume here (0.73). Paid close over 190.00 above average volume buys a starter into the 11/3 print; paid close under 168.25 on volume is a put. VRT prints 10/28 and it is the cleanest question on the board: 208.56, still below its 10-EMA at 209.41, 5.5% under the 20-EMA at 220.65, 13.6% under the 50-EMA at 241.45, yet +4.39% today on 1.44x, the 90th percentile of sixty sessions, RS63 improved from -25.9 to -6.5. I take no position into a binary twenty-four hours out. After it: a paid close over 209.41 that holds a day is the first honest evidence the buildout's financing cost has not eaten the equipment makers, and I take a starter call; a failed retest of 209.41 that rolls back under 200 puts me short the power leg toward 187.24. Off the sheet: AVGO at 342.33 closed under a falling 20-EMA at 342.99 on the only real volume in the basket — if this bounce had sponsorship, that is where it would have shown, and it did not. SMCI at 35.09 is my tell rather than my trade: the only name with up/down volume above 1.0 (1.38) after reclaiming 32.47, but 0.91x on a +7.3% day against a 20-EMA at 36.41 still reads like distribution's little brother. ORCL at 130.53, -0.47% in five sessions and 6.6% under its 50-EMA, is not a leader and does not get my capital. Seven names, one working long, one working hedge, five levels. Days are noise. I am positioned for the quarter.