When: Q3 FY27 earnings show continued DSO expansion with inventory write-downs

Then: Downgrade to Sell

When: 10%+ hyperscaler capex guide-down

Then: Downgrade to Sell

When: 30-year yield rollover below 4.5%

Then: Re-enter to Overweight

When: Q3 print shows accelerating sequential growth with constructive forward guide

Then: Re-enter to Overweight

When: Pullback to $180-190 with 200 SMA at $199.13 holding

Then: Re-enter with better risk/reward

SELL with tighter stops and explicit position-sizing discipline; staged exit beginning at $225 rather than waiting for $232.80; residual in cash/Treasuries

• Fundamentals: 75% gross margins, $96B revenue, 117% ROE — but receivables jumped 127% YoY to $63B while revenue grew 106%, a 21-percentage-point DSO extension
• Working capital deterioration: inventory grew 111% on 106% revenue growth — classic turn-deterioration pattern
• Balance sheet signal: management issued $25B in long-term debt to flip from net-cash to net-debt position; insiders hedging before cost of capital worsens
• Macro regime: 30-year at 23-year highs with Polymarket pricing 96% probability of zero Fed cuts in 2026 — no policy put underneath
• Valuation: 2.22 beta, 28.7x trailing earnings with forward multiple requiring near-perfect EPS doubling; a 30-50bp drift in 30-year breaks the math
• Recommended stop at $222 (below 10 EMA at $223.23), not $216 — the 50 SMA is a bounce launchpad, not a support floor
• Position sizing: cap at 60% of benchmark weight, with residual in cash or short-duration Treasuries earning 4%+ on the front end
• Staged exit: take 1/3 off near $225, another 1/3 toward $232, hold residual runner with $222 stop
• Execution risk: current $225 to $232.80 trim target is a 3.4% gap translating to 7.5% expected daily movement given 2.22 beta — gap-through risk is real
• Reject buy-the-dip at 50 SMA: in this regime, being early costs 20% drawdown before mean reversion; add on other side of rate reset or confirmed earnings beat

→ vs aggressive: Stop at $216 sits at the 50 SMA which is a bounce launchpad — will get clipped by dead-cat bounces and watch the stock rally away; push for $222 stop instead (4.8% giveback vs 10% giveback from current levels)

→ vs aggressive: Waiting for $232.80 trim target is discretionary and costs optionality — stock may gap through on rate-driven selloff before trim executes; lock in gains at current $225 level

→ vs neutral: 50 SMA is a mean-reversion level in a normal regime but this is not normal — buying a 2.22-beta name into 23-year highs in long-end yields with broken momentum and binary earnings risk costs 20% drawdown before mean reversion materializes

Trim to approximately 75% of benchmark weight; stage exit in thirds (one-third at $225, one-third at $230, final third held with $218.50 stop); park cash in short-duration Treasuries

• Reduce to ~75% of benchmark weight; do not exit; trend stack intact (price above all major moving averages with positive slopes)
• Exit trigger: confirmed close below 50 SMA at $215.79, not the $222 stop (which is only ~1 ATR below current price and will trigger on normal chop)
• Stage trim in thirds: first third at $225 (current area), second third at $230 (just below upper Bollinger band at $232.80), final third held as residual runner
• Stop for residual runner: $218.50 (~$6.50 below current, ~1.1 ATR), below September 14 swing low area but with buffer
• Custom silicon competitive threat operates on 24-to-36-month arc, not 3-to-6-month trader horizon; Anthropic's $7.33B compute spend up 190% YoY actually confirms demand durability
• September 29 relative-strength signal reflects institutional preference (NVDA bucking Dow selloff), not exhaustion—exhaustion would be failing to rally on a green tape
• Consensus EPS doubling requirement at 14.5x forward not disproven by latest quarter (revenue +105.9% YoY, net income +125.9% YoY, gross margin 75.0%)
• 60% weight cap is arbitrary; position size should derive from contribution to portfolio volatility, correlation with AI-exposed names, opportunity cost
• Re-entry triggers: 30-year yield rollover or earnings with raised guidance and maintained margin trajectory
• Tail risks (8% recession, 9% AI bubble) are meaningfully priced out, supporting trim (not exit) as appropriately calibrated response
• Park cash in short-duration Treasuries (not money market); 2.22 beta means residual NVDA position does more work than its weight suggests

→ vs aggressive: September 29 relative-strength reading is institutional preference, not exhaustion—exhaustion would be failing to rally on a green tape; aggressive analyst flipped a bullish signal into bearish by rhetorical force

→ vs aggressive: Forward P/E 'consensus trap' argument logically incomplete; latest quarter shows acceleration (revenue +17.9% QoQ, +105.9% YoY; gross margin expanding 72.4% to 75.0%), not deceleration signature

→ vs aggressive: Custom silicon thesis has timescale mismatch—TPU, Trainium, Maia operate on 24-to-36-month arc; trader's 3-to-6-month horizon cannot support structural competitive argument; Anthropic $7.33B compute spend up 190% YoY cannot simultaneously argue demand durability and competitive threat

→ vs conservative: $222 stop is reactive and insufficient—ATR is $5.86; $222 is less than 1 ATR below current price and barely below 10 EMA at $223.23; will trigger on any single-session 1.5% move in a 2.22 beta name

→ vs conservative: $222 stop abandons structural reference entirely; 50 SMA at $215.79 was tested as support September 14 and held with memory; aggressive analyst's $216 stop sits appropriately at that structural level

→ vs conservative: 60% benchmark weight cap is arbitrary with no derivation from risk model or volatility budget; position size should be function of contribution to portfolio volatility, correlation with AI names, opportunity cost

→ vs conservative: Executing at $225 instead of $232.80 leaves 3.4% on table; contradicts 'trim on strength' concept while simultaneously recommending staged exit—cannot recommend both staged execution and front-load at lower price