When: Confirmed close >$525 after tier 1 re-entry confirmed

Then: Add additional 25% position to re-establish full size

SELL - Trim 50 to 65% immediately at current levels; stop on residual at 410; re-entry only on 50 SMA reclaim + MACD histogram crossover

• A 25 to 40 percent trim leaving 60 to 75 percent residual is not genuine risk management—approximately 18 to 22 percent of a 30 percent drawdown exposure remains
• Execution plan assuming a relief rally to 489 to 490 is fragile; if AMD goes straight to 438 without tagging 489, the plan fails
• The 438 stop is dangerously tight: the 14-day ATR is 29.39, and the July 29 swing low was 429—a single oversold wick can trigger it
• A volatility-adjusted stop should be closer to 410 to 411 (2 ATRs below entry), not 438
• Google committed 120 billion to TPU and added Marvell for 12.2 billion; Waymo built a 5-nanometer ASIC; Meta, Amazon, Microsoft all building custom silicon
• AMD's premium is priced for AI inference optionality that is being structurally compressed quarter by quarter
• The macro data (86 percent odds of zero Fed cuts in 2026, rising 10-year yields, Walmart consumer softness, 35 percent Japan recession odds) supports deeper reduction
• ARK is selling AMD and buying Nvidia; retail is asking how to trim—bull case is carried by opinion pieces, not flow
• Counter-proposal: trim 50 to 65 percent now, stop at 410, re-entry trigger requires both 50 SMA reclaim on closing basis AND positive MACD histogram crossover

→ vs aggressive: A 25 to 40 percent trim is a half-measure that protects sentiment about the AI thesis while accepting the bulk of the downside—residual exposure of 60 to 75 percent is not conservative risk management for a stock with beta of 2.49 at 30 times forward earnings and nearly 120 times trailing multiple

→ vs aggressive: Waiting for a relief rally to 489 to 490 assumes the bounce materializes; if AMD goes straight to 438 without ever tagging 489, the trim plan is never executed—hope is not a risk management tool

→ vs aggressive: The 438 stop is just above the July 29 swing low of 429; with daily ranges exceeding 50 dollars during the July breakdown and 14-day ATR at 29.39, a single oversold spike wick can tag the stop, lock in losses, and force exit from the core position—this is a coin flip, not a stop

→ vs aggressive: Google committed 120 billion to TPU, added Marvell for 12.2 billion in custom inference silicon, Waymo built its own 5-nanometer ASIC, Meta has Trainium, Amazon has Trainium and Inferentia, Microsoft has MAIA—every major AMD customer is engineering around the merchant accelerator model

→ vs aggressive: The premium at 766 billion dollars market cap cannot be justified by EPYC, gaming, and embedded segments—it's priced for AI inference optionality that is being structurally compressed; trimming 60 to 75 percent is more rational than 25 to 40 percent

→ vs aggressive: Using the same macro data (86 percent odds of zero Fed cuts in 2026, rising 10-year yields, Walmart consumer softness, 35 percent Japan recession odds), the appropriate response is deeper reduction, not a modest trim—the asymmetry cuts against high-multiple high-beta names

→ vs aggressive: ARK selling AMD and buying Nvidia, retail asking how to trim, and both ends of the market quietly distributing suggests the bull case is carried by opinion pieces and price-target speculation, not by flow—when marginal buyer disappears, hold a smaller residual, not 60 to 75 percent

HOLD (Trim 40-50% of position: half at market ~469, half via limit orders 482-490; residual 50-60% with stop at 420, re-entry tiered at 500 and 525)

• Lower-high sequence from 580 to 514 to 469 represents real distribution evidence
• Macro backdrop is hostile with 86% no-cut odds, rising real yields, and Walmart consumer softness
• AMD fundamentals are strong: 50% revenue growth YoY, 156% EPS growth, operating margins expanded from negative to 17.3% in four quarters
• AMD has $13.1B cash and short-term investments against $3.2B financial debt, generating $8.8B free cash flow on trailing twelve-month basis
• Custom-silicon/inference ASIC competitive threat is multi-year process, not quarter-over-quarter cliff
• ATR of 29.39 requires minimum two ATRs below entry for appropriate stop placement (~410-411 minimum, analyst uses 420)
• Current MACD histogram at minus 0.62 with 50 SMA sloping down; re-entry requires close above 500 with positive MACD histogram and RSI above 50
• July swing low at 429 and Boltzmann lower band at 440 support 420 stop level as genuine trend damage threshold
• Custom-silicon/competitive erosion is slow-burn structural issue, not imminent cliff warranting deeper cuts

→ vs aggressive: Calling a 25-40% trim 'aggressive' is rhetorical inflation; leaving 60-75% exposed is not preserving the secular thesis when marginal buyer is gone and macro is hostile

→ vs aggressive: Waiting for relief rally at 489-490 is discretionary trading dressed as technical discipline; VWMA sloping down with lower-high sequence locked in means no statistical obligation for price to revisit that level

→ vs conservative: Selling 50-65% immediately at 469 is selling into weakness; same trim at 490 is four to five percent better execution with no additional risk

→ vs conservative: Requiring both 50 SMA reclaim AND positive MACD histogram for re-entry is double-barreled and forces buying the breakout rather than the base at significant cost