When: Close below $196 (200 SMA break)

Then: Hard invalidation of thesis

HOLD; trim into strength at $219 to $225; do not add new capital until confirmed base or macro pivot

• PEG of 0.46 is a snapshot, not a floor—drops to 0.42 at $190 and 0.38 at $171, but cheapness evaporates if growth disappoints
• 2022 NVIDIA drawdown from ~$33 to ~$11 (67%) saw MACD bearish entire way down, 200 SMA rising while stock halved—survivorship bias in bullish pattern argument
• Polymarket probabilities: 72% chance of touching $208 vs 8% chance of touching $240; 94% probability of no rate cuts in 2026; expected low is $200-$215
• Beta of 2.2 means 10% broader market drawdown translates to ~22% on NVDA; 10-year just touched 5%; recession odds ticked 7% to 10% in one week
• Custom silicon threat is broader than Cornelis: Google TPUs, Amazon Trainium/Inferentia, Microsoft Maia, Meta MTIA—all hyperscalers reducing NVIDIA dependence
• Receivables balloon from $38B at fiscal year-end to $63B in six months (64% increase vs ~40% revenue growth); days sales outstanding stretching
• Collar capping upside at 17% and protecting against 9% downside is capital preservation, not return maximization—appropriate for fiduciary obligations during binary event risk
• Current setup is highest-uncertainty macro tape in two years: hostile macro, bearish prediction markets, sector rotation fleeing chips, Fed meeting imminent

→ vs aggressive: PEG of 0.46 is not a floor; it's a one-dimensional snapshot that ignores growth uncertainty from hyperscaler efficiency gains, custom silicon acceleration, and AI exec slowdown calls

→ vs aggressive: Every-prior-bottom pattern argument has survivorship bias; 2022 saw 67% drawdown with MACD bearish entire way down and 200 SMA failing as lagging indicator

→ vs aggressive: Cannot cherry-pick right-tail scenarios (30-50% upside) while dismissing left-tail as 'only 5%'; Polymarket implies modestly negative expected value with fat left tail

→ vs aggressive: Collar critique misframes risk management as 'mathematically wrong trade'; capital preservation during binary event risk is fiduciary responsibility, not defensive posture

→ vs aggressive: Dismissed Cornelis ($205M) but ignored broader custom silicon programs from Google, Amazon, Microsoft, and Meta—selection bias, not comprehensive competitive analysis

→ vs aggressive: Ignored receivables growth of 64% vs 40% revenue growth and stretching days sales outstanding—cherry-picked fundamentals report for PEG and FCF while skipping balance sheet deterioration signals

HOLD with refined trade management: trim 25-35% into rallies toward $212-$225, defer collar decision until post-Fed meeting, and establish conditional add-back/exit triggers based on price action.

• Fundamentals are elite with PEG of 0.46 and forward PE of 13.5 for a company growing this fast
• Macro regime is hostile: yields breaking 5%, Fed showing 94% probability of no cuts in 2026, recession odds ticking up 3 percentage points in one week
• Conservative analyst correctly flagged receivables build (64% growth vs 40% revenue growth in six months) but overstates it—customers are hyperscalers and sovereign AI buyers, not channel distributors
• Aggressive analyst incorrectly uses 200 SMA as structural floor when it failed from 2021 highs, with stock halving over eight months
• Networking moat through Mellanox and Spectrum-X creates system-level lock-in that makes custom silicon a complement rather than substitute
• Collar is overpriced at current implied volatility; defer hedge until after Fed meeting when vol likely lower
• Decision should be path-dependent: expected value calculation changes based on whether $200 support holds or breaks
• Trim zone should increase from 15-25% to 25-35% given risk asymmetry—path to $200 is only 5% away, path to $230 requires Fed pivot or major AI sentiment reversal

→ vs aggressive: Treating fundamentals as a substitute for timing—200 SMA failed as floor from 2021 highs and stock halved over eight months; receivables build of 64% growth vs 40% revenue growth was ignored; not buying a setup, underwriting a regime

→ vs conservative: Receivables concern overstated—customer composition is hyperscalers and sovereign AI buyers, not channel distributors, so DSO stretch is likely due to revenue mix shift toward enterprise deals; probability framework is internally inconsistent by accepting Polymarket pricing of 8% upside without interrogating why the market is pricing it that way