When a memory-chip giant commits nearly $29 billion to buying back its own stock, it's management signaling confidence that chip prices and demand will stay strong — they'd never spend that money if they believed the cycle was topping. Meanwhile, the AI i

Verdict: a real memory cycle, but the entry isn't armed — wait for the reclaim

**Verdict: wait — the cycle is real, the entry is not.** The strongest evidence for this trade is that Micron's fundamentals have inflected violently in the thesis's favor: quarterly revenue ran from $8.7B in the quarter ending November 2024 to $41.5B in the quarter ending May 2026, gross margin hit 84.6% that quarter, and the 60-month backtest returned 239.9% with the gains concentrated from mid-2025 onward — exactly when the inference-demand story showed up in the numbers. The strongest evidence against it is that memory is brutally cyclical: the same company printed negative gross margins and burned over $6B of free cash flow in fiscal 2023, the backtest endured a 32.9% drawdown on just 6 trades with a 50% win rate, and MU and WDC are 70% correlated, so this basket is roughly 1.3 independent bets, not two. Live conditions confirm the wait: MU's RSI (14) sits at 45.4 versus the required cross above 50, and WDC trades about $70 below its 50-day EMA, so neither entry is armed. With quarterly earnings from both names due within weeks, the highest-information events are still ahead. **Conviction breakdown:** Thesis support 72 — the buyback signal and inference demand are corroborated by margins, but the SK hynix financials behind the $28.6B buyback could not be cross-checked here. Trade readiness 40 — two of six entry conditions met on MU, none close on WDC. Risk quality 45 — a 32.9% realized drawdown, tight 2.5% stops in ~66-71% volatility names, and daily-bar exit fills make reported risk approximate. Backtest evidence 60 — a strong 239.9% return over 60 months, but only 6 trades, a 50% win rate, and a 24-month window that won just 1 of 4. Fundamentals trend 88 — revenue, margins, free cash flow, and deleveraging (debt-to-equity down to roughly 0.05 for MU) are all moving the right way.

Trade now

**MU (last close $910.52) — the entry is not fully in play yet.** The strategy is a "Confirmed 50-Day Reclaim Long": price must touch or dip below the 50-day EMA ($907.27) intraday, then close back above it on the same bar, with RSI (14) crossing above 50 and ADX (14) above 25. Right now MU sits $3.25 above its 50-day EMA, RSI is 45.4 (it needs to push above 50 — roughly 4.6 points short), and ADX is 25.1, only narrowly clearing its bar. Two of six conditions are met (RSI under 65, ADX above 25); the reclaim-and-cross sequence is not. **Wait** means just that: stand aside until a bar where MU tags the EMA near $907 and closes above it while RSI pushes through 50. **WDC (last close $435.39) — same playbook, much deeper discount.** WDC trades $70.06 below its 50-day EMA ($505.46), so the reclaim condition is nowhere near firing, even though ADX at 38.9 comfortably beats 25 and RSI at 35.7 sits well under the 65 ceiling. WDC would need roughly a 16% rally just to get back to its EMA — treat this as a reversion setup in waiting, not a live entry. **Risk framing if an entry does trigger.** The strategy's hard stops are a 2.5% fixed-risk stop and the first support level; on the 60-month backtest this rule set produced 6 trades, a 50% win rate, and a 240% return, but with a 32.9% peak-to-trough drawdown — which is exactly why position sizing caps at 25% of capital per position. Exits come at a 5% fixed take-profit, a Fibonacci extension target, or RSI above 75 with price above the upper Bollinger band. One caveat on the fills: exits were judged on daily bars, not intraday, so reported trade quality is only approximate.

Why the bull case has support

The thesis rests on two pillars — a capital-return signal from a memory incumbent and an AI inference demand wave — and the fundamentals back both. SK hynix's $28.6 billion buyback announcement (per the Yahoo Finance piece of August 25, 2026) is the trigger, and the idea argues management would not commit that capital if it saw the cycle topping. The companion Yahoo piece on Broadcom versus Micron makes the demand case explicitly: capital is shifting from training to inference and agentic AI, and inference is unusually memory-hungry — exactly Micron's product. Micron's reported numbers show the cycle inflecting hard. FY2025 revenue grew 11.5% year over year to $37.4 billion, but the quarterly trajectory is the real story: revenue ran from $8.7 billion in the quarter ending November 2024 to $41.5 billion in the quarter ending May 2026 — a near-quintupling in six quarters. That is not a steady-eddie business; that is a memory price cycle in full swing, and it puts Micron in the 71st percentile of Information Technology…

Scores

• Conviction score breakdown: 61
• Thesis support: 72
• Trade readiness: 40
• Risk quality: 45
• Backtest evidence: 60
• Fundamentals trend: 88

Watch items

• MU — Close vs 50-day EMA (reclaim: dip to or below EMA, close above)
• MU — RSI (14)
• MU — ADX (14)
• WDC — Close vs 50-day EMA
• WDC — RSI (14)
• MU — Nearest support
• WDC — Nearest support
• MU — Price
• MU — Price above EMA (50)
• MU — Price crossed above EMA (50)
• MU — RSI (14) crossed above 50
• MU — RSI (14) below 65
• MU — ADX (14) above 25