HOLD with no incremental sizing

• Stock has 2.22 beta with historical memory cycle drawdowns of 40-60% over months
• Sell-side consensus has been wrong by 30-60% at cycle tops (fiscal 2022 expected ~$8-9 EPS, actual fiscal 2023 was -$5.34)
• Capex running at $25 billion/year with construction in progress doubling from $5.5B to $10.9B signals late-cycle positioning
• Operating margin swings from peak to trough historically 60-80 percentage points
• Receivables of $31 billion on $90 billion revenue base equals ~124 days sales outstanding, vulnerable to hyperscaler payment term changes
• Receivables growing at 234% YoY creates working capital risk
• Aggressive entry at $1,065 with stop at $971 risks 8.8% to make 5.3% to $1,121 or 12.7% to $1,200
• Conservative waiting strategy targets $1,005 entry with stop below $900
• Daily ATR of 4.4% makes ATR-based stops prone to being run
• 30-year Treasury at highest level since 2002, mortgage rates at 7.58%
• Polymarket pricing 96% probability of no Fed cuts in 2026
• Structural thesis is intact but fully reflected in current price, leaving no margin of safety for errors

→ vs aggressive: Trend-following indicators (price above 10 EMA, 50 SMA, 200 SMA, positive MACD, neutral-bullish RSI) do not protect against 40-60% memory cycle drawdowns; trend-following into final innings is a value trap dressed in momentum clothing

→ vs aggressive: Forward PE of 6.6x anchored on $161 consensus EPS is a marketing projection, not a floor; consensus has been wrong by 30-60% at every prior cycle top

→ vs aggressive: Capex at peak margins is the most reliable late-cycle signal; fiscal 2022 capex of $12B followed by fiscal 2023 loss of $5.34/share proves over-building at the top destroys value

→ vs aggressive: Receivables of $31B representing ~124 days of sales is vulnerable if any hyperscaler extends payment terms by 30 days during digestion period

→ vs aggressive: The recovery from $820 to $1,065 is survivor bias; macro headwinds (7.58% mortgage rates, highest 30-year Treasury since 2002, 96% probability of no 2026 Fed cuts) are persistent, not binary events resolved by bounces

→ vs aggressive: Cost of waiting vs. cost of being wrong: risking 8.8% to make 4-13% is sub-one risk-reward on near-term move; patient trader waiting for $1,005 catches 10% lower entry with better stop

→ vs aggressive: WSB call spread popularity is a contrarian indicator, not bullish; retail paying high premiums means market makers are selling that upside, already in the price

Hold core position and add smaller tactical increment of 10-15% in two tranches with tighter stops than aggressive proposal; rebalance broader exposure if MU is concentrated

• Technical setup is constructive: price above all major MAs, MACD positive and expanding, RSI neutral-bullish, VWMA confirming volume
• Cycle risk is real: peak operating margins in a 2.22 beta memory name with multi-cycle history of 40-60% drawdowns
• Consensus EPS at top of memory cycles has been wrong by 30-60% historically; treat consensus as optimistic case, not floor
• Moderate path: 10-15% tactical increment (not 30-40% aggressive proposal) in two tranches
• First tranche at ~1,065 with stop at 990-1,000 (below Bollinger middle band at 1,005)
• Second tranche conditional: either pullback holding 10 EMA at 1,044 with RSI >50 and MACD histogram positive, OR breakout above 1,108 on volume >25M shares
• Core position stop at 920-930 rather than aggressive 971 (50 SMA retest to 950 is normal, not breakdown)
• Restrictive macro policy is a tax on multiple, not collapse trigger; accept 950-1,150 range through year-end
• Buy weakness within range, sell strength; avoid chasing breakouts or hiding in cash
• MU at peak margins with 2.22 beta should not be largest portfolio position; concentration risk is a real factor

→ vs aggressive: Consensus EPS of $161 is sell-side optimism at cycle peak, not a floor; actual outcomes land 30-60% below consensus

→ vs aggressive: Sub-one risk-reward ratio on near-term move (8.8% downside vs 5% to 1,108 and 13% to 1,200) is not asymmetric; adding full size at midpoint with wide stop is poor risk management

→ vs aggressive: Trend-following at memory cycle peaks has historically been one of most reliable ways to lose money; stacked MAs, positive MACD, and rising VWMA were all present at June peak of 1,255 preceding drawdown to 820

→ vs conservative: Prior memory cycles are not a direct template; structural setup differs meaningfully with 3-supplier HBM oligopoly, hyperscaler AI capex funded by operating cash flow, and HBM4 gating Nvidia allocations

→ vs conservative: Receivables DSO of 124 days is a watch item, not a deal-breaker; what matters is customer concentration and payment incentives, not absolute DSO

→ vs conservative: Macro headwind being both priced in (to justify adding) and not priced in (to justify holding) is inconsistent survivor bias

→ vs conservative: WSB call spread is not just noise; sophisticated retail positioning at specific strikes provides forward-looking sentiment signal that should be weighted, not dismissed