$MU Micron, still in the bookMicron Technology, Inc.BATS:MUEXCAVOGross margin 84.6%. Holding since October 31, 2025. Entered at 223.54, yesterday's close 1082.01. That's +384% on a position that has been through eleven monthly rebalances. A boring cyclical name, the kind people buy for a DRAM (Dynamic Random-Access Memory) upturn and dump two quarters later, just started printing money like a mint. That's why my hand doesn't reach for the sell button. The idea Micron became a clean bet on the HBM (High Bandwidth Memory) shortage, the exact memory that AI (Artificial Intelligence) accelerators cannot run without right now. Everyone bets on who builds the AI. I bet on the one whose component the AI won't even boot without, and who has already locked in $100B+ of demand ahead of time. The difference is simple. One camp sells a dream at a floating price. The other sells hardware on signed contracts. I wrote about this in my last post about robots and the hands for those robots. Last night I watched the film Upgrade from 2018, and that is exactly what it's about. How the position was built The mechanics matter here. Micron wasn't bought once and forgotten. The portfolio is built from three layers offset in time, each reviewed on its own quarterly cycle. Micron got reselected in every layer, most recently in June, July and August 2026. Right now it occupies all three layers at once, which is the maximum weight one name can carry in this book, 6.67%. Eleven monthly rebalances have passed, and the position never lost its seat once. On the first of every month the portfolio rebalances by algorithm. Micron can be sold, added to, or partially closed, depending on what the system says. We find out at the start of the next month. The algorithm doesn't know the position is up +384%. It only knows what it sees in the data right now. Why now: the numbers don't lie Micron closed its latest fiscal quarter at records. Revenue around $41.46B, an all-time high, gross margin 84.6%, adjusted EPS (Earnings Per Share) around $24.67. A year ago the same company was barely climbing out of losses. Now it prints profit on every report. The model broke, in a new direction. Guidance for the next fiscal quarter: revenue around $50B ±$1B, margin near 86%, EPS around $31. The company is raising its own bar, as if it's in a hurry. But watch the pace. Off a record quarter, the guidance implies about +10% QoQ (quarter over quarter), and the acceleration itself is starting to fade. Remember that number. It leads straight to risk number one. HBM4 12-high (fourth-generation HBM, a 12-die stack) is ramping roughly twice as fast as HBM3E (the previous generation, HBM3 Enhanced) did in its day. Micron has already shipped over $1B in HBM4 revenue. And the key part, $100B+ in customer contract commitments locked in ahead. Demand is contracted in advance, unlike spot, where the price evaporates overnight. The mechanism: where the catch is Most people look at a margin near 85% and reflexively shout "cycle top." In the usual memory story that would be right. Margins like that always pulled in capital, everyone built fabs, overproduction kicked in, and a year later margins hit the floor. But here the old mechanic breaks in one spot. HBM doesn't get made on command. Ordinary DRAM reaches volume in a handful of quarters, and this is different. On top of that, those $100B of contracts mean demand is locked in ahead and holds regardless of how spot feels next week. A customer who signed for deliveries doesn't dissolve if the AI hype cools a little. That's why analysts, for the first time in a decade, are talking about an "extended cycle" instead of the usual cliff. Coverage is clean: 38 analysts, 4 Strong Buy, 31 Buy, 3 Hold and zero Sell. Exactly three skeptics, countable on one hand. As long as the contracts hold and the fabs physically can't catch up to demand, the extended-cycle mechanic works. That's the case for holding. Structure and levels: where the market stands today Price is 13.8% below the high of 1254.52, holding above the 20-day at 993.99 and the 50-day at 941.35. The 200-day is far below at 660.85. The last significant low is 770.42. ATR(14) (Average True Range) is 44.21, about 4% of price. The daily range here is wide, and you have to treat it that way. Nearest resistance sits barely half a percent above: 1088.98, the June 3 swing high. Price is testing it from below right now. This is the main fork for the whole idea. Consolidation above 1088.98 opens the road to the all-time high at 1254.52. A rejection there leaves price stuck in the range. Two scenarios, and the gap between them is under a percent. The support picture is layered. First real one is 993.99 (20-day, -8.1% from spot). Below it 941.35 (50-day, -13.0%) and almost right against it 936.27, the main volume node of the past six months, 12 million shares. That's a cluster about 2.5% wide, and you draw it as a rectangle. A single line understates it. Next tier: 919.11/912.70 (-15%), the September 3 swing low plus a second volume node. Then 902.76/888.00 (-17%), the September 14 and August 24 swings. Below 888 it's empty down to 770.42. Bare space in between: no swings, no volume. If the 936-912 cluster doesn't hold, the drop will be fast, because there is nothing there to slow it. Losing the 20-day, then the 50-day, would be the first real deterioration of the structure that has held since last autumn. Below the September low of 770.42, the trend that made this whole move stops being intact. What could break this First, numbers that are too good. Remember the +10% QoQ from the guidance? That's the first crack. Triumphant year-over-year comparisons are physically impossible to repeat forever. Next year's base will be brutal. When revenue lands "merely" flat, the ticker reads it as a slowdown. Second, overproduction. An industry with margins near 85% is a magnet for capital. Samsung and SK Hynix won't watch this feast from the sidelines. HBM is protected by complexity, but that protection is temporary. Around $27B of capex (capital expenditure) for the fiscal year is armor and fuel at the same time. Armor: HBM capacity keeps Micron irreplaceable today. Fuel: the whole industry's capex today becomes someone else's fabs in two years. What protects the margin now is preparing its cliff later. Third, unanimity as a contrarian signal. 31 Buy and zero Sell is a reason to hold and a red flag at the same time. When the entire analyst book stands on one side, there's no one left to disappoint. Every upgrade is already priced, and the first downgrade hits empty air. When stock trackers pump targets on Twitter and retail chatter turns "up only," it's already late in the evening. Whoever walked onto the dance floor last pays the highest cover. Fourth, rates. The Fed (Federal Reserve) is still holding a hard line, and the market is pricing a hold instead of gentle easing. Expensive money weighs on the multiples of expensive growth stories hardest of all. Fifth, "sell the news." A stock that falls on a record report has already told you everything. The appetite for news is spent. A drop on good numbers is my personal trigger to take some off, no arguing with the tape. Overproduction in memory is inevitable. It always comes. But between "inevitable" and "now" there is room for several more record quarters. The account can die before the inevitability reaches the P&L (Profit and Loss). The only question is which side of the trade you're standing on when the music stops. What's next On the first of every month the algorithm reviews all three portfolio layers. Micron can stay at max weight, can be partially closed, can be added to. The system decides on the data. Emotion doesn't get a vote. Neither you nor I know in advance what it will say on the 1st. That is exactly how it's supposed to work. Eleven months, eleven rebalances, +384%. The thesis still holds. While it holds, I hold. Best Regards, EXCAVO