Here’s the bear case (and strategy): Anyone else amazed at the content generation that these programs can provide? Again, the devil is in the details. As you can see below, there are several market timing events that must be right. With portfolio values on the line, it rhymes, but does the song sound like the mood you’re in? Interesting considerations, however.
WARNING: Short selling and other highly controversial approaches here. Classically, this is the definition of committing to/against a thesis. Going to cash/bonds/stable income is a feature here, however.
Scenario 3: Retreat to Higher Ground
Core Thesis
AI infrastructure valuations have reached speculative extremes reminiscent of prior technology bubbles (1999-2000, 2021). Creative destruction dynamics—where new technologies rapidly displace incumbents—combined with commoditization pressure will lead to sustained value destruction. The rational strategy is capital preservation with optionality for re-entry at more attractive valuations.
Supporting Arguments (Bear Case for AI Infrastructure)
-
Valuation Extremes: NVIDIA at 45x earnings with 75% gross margins is unsustainable; historical parallels (Cisco 2000) are ominous
-
Commoditization Inevitable: GPUs face ASIC competition; cloud services face margin pressure; AI features become table stakes
-
Demand Uncertainty: Enterprise AI ROI remains unproven; 2026-2027 could see investment pause as businesses reassess
-
Overbuilding Risk: $500B+ in AI infrastructure investment 2023-2025 may exceed demand by 2-3 years
-
Technology Disruption: Post-transformer architectures, neuromorphic computing, or quantum breakthroughs could obsolete current infrastructure
-
Regulatory Pressure: Antitrust actions against hyperscalers, energy/environmental constraints, export controls create headwinds
-
Macro Vulnerability: Higher-for-longer interest rates disproportionately hurt high-multiple, capital-intensive businesses
Historical Parallels
Dot-Com Bubble (1999-2002):
-
Cisco peaked at 35x sales, fell 89% over 2.5 years
-
Qualcomm fell 80% despite fundamental business strength
-
Recovery to peak valuations took 15+ years
Cloud Infrastructure Bubble (2021-2022):
-
Cloud stocks (SNOW, NET, DDOG) fell 60-75% despite strong growth
-
Even AMZN and MSFT fell 50% and 35% respectively
-
Taught lesson: High growth doesn’t immunize against multiple compression
Current Situation:
-
AI infrastructure stocks up 150-300% since 2023
-
Valuations at/above prior peaks on more aggressive assumptions
-
Institutional ownership at record levels (over-consensus risk)
Year-by-Year Evolution (2025-2030)
2025-2026: The Reckoning Begins
-
GPU oversupply emerges as Blackwell ramp exceeds demand; NVDA falls 35-45%
-
Enterprise AI spending growth disappoints (30% vs. 50% expected); cloud multiple compression
-
First major data center write-downs as demand projections revised
-
Utilities face power demand shortfall; stock prices decline 20-30%
-
Market capitulation in AI infrastructure, sector underperforms by 25-35%
-
Expected Portfolio Return: +3-8% (preservation relative to -25% sector decline)
2027-2028: Creative Destruction Accelerates
-
ASIC performance parity achieved; NVIDIA margins compress to 55-60%
-
Cloud commoditization intensifies; AWS/Azure pricing pressure
-
Data center oversupply reaches 20-25%; occupancy rates fall, rents decline
-
Next-generation architecture creates winner/loser bifurcation
-
Survivor bias evident: A few winners, many losers
-
Expected Portfolio Return: +2-5% (preservation + selective re-entry)
2029-2030: New Cycle Foundation
-
AI infrastructure resets to sustainable growth rates (12-15% annually)
-
Valuations reach trough multiples (20-25x for quality assets)
-
Clear signals emerge for next-generation winners
-
Strategic re-entry into select positions at 50-70% discounts to 2025 peaks
-
Expected Portfolio Return: +8-15% (early reinvestment captures recovery)
Portfolio Construction
Objective: Capital preservation with 0-5% real returns, maintaining optionality for aggressive re-deployment when risk/reward becomes attractive
Phase 1 (2025-2026): Defensive Positioning - 70% Portfolio
-
Short-Duration Treasury Fund (SHY, VGSH): 30%
-
Investment-Grade Corporate Bonds (LQD, VCIT): 15%
-
Dividend Aristocrats (NOBL, VIG): 15%
-
Rationale: Defensive sectors (consumer staples, healthcare), 2.5-3.0% yields
-
Lower correlation to tech volatility
-
Gold & Commodities (GLD, DBC): 10%
Phase 1 (2025-2026): Strategic Short Positions - 20% Portfolio
-
Inverse Semiconductor ETF (SOXS - 3x Inverse): 8%
-
Rationale: Profit from GPU bubble deflation
-
Target: NVDA falls from $485 to $280-320 range
-
Stop Loss: If NVDA breaks above $550 (thesis invalidated)
-
Expected Return: +60-90% if thesis plays out
-
Put Options on NVDA (12-18 month duration): 5%
-
Strike: $400-420 (20% out of money)
-
Rationale: Asymmetric downside capture with limited risk
-
Cost: ~3-4% of notional, decay acceptable given thesis
-
Target: Triple if NVDA falls below $350
-
Put Options on CLOU (Cloud ETF): 4%
-
Strike: 15-20% below current levels
-
Rationale: Cloud multiple compression trade
-
Duration: 12 months, roll if necessary
-
Short Position in Speculative AI Stocks: 3%
-
Targets: High-multiple, unprofitable AI application companies
-
Examples: Select names trading >20x revenue with negative cash flow
-
Stop Loss: -20% on individual positions
Phase 1: Cash Reserves - 10% Portfolio
-
Dry powder for Phase 2 re-entry
-
Maintain flexibility for tactical opportunities
-
Psychological comfort during volatility
Phase 1 Trigger Points (When to Exit Defensive Posture)
Green Lights for Phase 2 Transition (Need 3 of 5):
-
NVDA trading below $320 (35% decline from Sept 2025)
-
Nasdaq-100 decline of 25%+ from peak
-
Data center REIT FFO multiples compress to <18x
-
GPU delivery times normalize to <2 months (oversupply signal)
-
Enterprise AI spending growth decelerates to <20% YoY
Red Lights (Thesis Invalidation - Exit Shorts Immediately):
-
NVDA announces breakthrough that extends moat (e.g., 5x performance leap)
-
Enterprise AI ROI data proves compelling (>30% productivity gains)
-
New use cases drive demand inflection (e.g., humanoid robots, autonomous vehicles scale)
-
Regulatory tailwinds (e.g., government mandates accelerate AI adoption)
Phase 2 (2027-2028): Selective Re-Entry - Portfolio Transitions to 40% Equities
Timing: Initiated after 25-35% AI infrastructure correction, valuations reset to historical norms
Phase 2 Core Holdings (40% of portfolio):
-
NVIDIA (NVDA): 8%
-
Entry Target: $280-320 (down from $485)
-
Thesis: Survivors’ moat strengthens post-shakeout; margin stabilization at 60-65%
-
Position Size: Much smaller than Scenario 2 (8% vs. 20%) - conviction rebuild phase
-
Microsoft (MSFT): 8%
-
Entry Target: $320-350 (down from $425)
-
Thesis: Azure + enterprise software diversification provides downside protection
-
Best-of-breed survivor with balance sheet fortress
-
Constellation Energy (CEG): 6%
-
Entry Target: $180-200 (down from $245)
-
Thesis: Nuclear power infrastructure still scarce; long-term PPAs provide floor
-
Defensive infrastructure play with growth optionality
-
Amazon (AMZN): 6%
-
Entry Target: $135-150 (down from $178)
-
Thesis: AWS + retail diversification; cash flow machine at attractive valuation
-
Re-entry after market capitulation
-
Digital Realty (DLR): 5%
-
Entry Target: $110-125 (down from $155)
-
Thesis: Quality data center infrastructure at distressed valuations
-
5-6% dividend yield provides margin of safety
-
AMD (AMD): 4%
-
Entry Target: $100-115 (down from $165)
-
Thesis: NVIDIA alternative at discount pricing; market share gains
-
Higher risk/reward than NVDA
-
Oracle (ORCL): 3%
Phase 2 Continued Defense (50% of portfolio):
-
Treasury & Bond Allocation: 35%
-
Reduced from 45% but still substantial
-
Yields likely 5-6% if Fed responds to market stress
-
Maintain liquidity for Phase 3 acceleration
-
Dividend Aristocrats: 10%
-
Gold/Commodities: 5%
Phase 2 Cash: 10%
Phase 2 Risk Management
Position Sizing Discipline:
-
Maximum single position: 8% (vs. 20% in Scenario 2)
-
Conservative rebuild reflects uncertainty about bottom
-
Dollar-cost average entries over 3-6 months
Stop Losses:
-
Looser than Scenario 2 (accepting 25-30% volatility)
-
Focus on thesis breaks, not price action
-
Willing to endure “dead cat bounce” fakeouts
Expected Phase 2 Outcome:
-
Portfolio protection during Phase 1 limits drawdown to -10 to -15%
-
Phase 2 re-entry captures 50-60% of recovery
-
Total return 2025-2028: +15-25% (vs. -20 to -30% for buy-and-hold)
Phase 3 (2029-2030): Aggressive Redeployment - Portfolio Reaches 70% Equities
Timing: Clear signals emerge that AI infrastructure has found sustainable foundation; next-generation winners identifiable
Phase 3 Thesis: The correction has cleared speculative excess. Valuations are attractive. Technology roadmaps are clearer. Enterprise AI ROI is proven. Time to shift from preservation to wealth creation.
Phase 3 Portfolio Allocation (70% Equities, 30% Bonds/Cash):
Tier 1: Proven Survivors (40% of portfolio):
-
NVIDIA (NVDA): 12%
-
Accumulated position from Phase 2 entry ($280-320) now at $450-500
-
Thesis Validated: Software moat proved durable; margins stabilized at 62-65%
-
Hold for continued recovery to $650-700
-
Microsoft (MSFT): 10%
-
Amazon (AMZN): 8%
-
Constellation Energy (CEG): 5%
-
Entry at $180-200, now at $280-300
-
Nuclear power premium validated; long-term AI infrastructure winner
-
Hold for $400+ as power scarcity resurfaces
-
Digital Realty (DLR): 5%
-
Entry at $110-125, now at $165-180
-
Data center recovery; occupancy rates climbing back to 85%+
-
FFO growth resuming at 8-10% annually
Tier 2: Next-Generation Winners (20% of portfolio):
-
Emerging AI Infrastructure Leader: 7%
-
Example: Next-gen inference company (e.g., Groq if public, or equivalent)
-
Thesis: Post-shakeout winner in inference specialization
-
High conviction in technological differentiation
-
Vertical AI Application Leader: 7%
-
Example: Healthcare AI platform with clear ROI and defensible moat
-
Thesis: Application layer value capture begins in earnest 2029-2030
-
Category leader with network effects
-
Edge AI/Distributed Compute Play: 6%
-
Example: Edge data center or on-device AI leader
-
Thesis: Architectural shift toward distributed inference
-
Early positioning in next wave
Tier 3: Cyclical Recovery (10% of portfolio):
-
AMD (AMD): 5%
-
Entry at $100-115, now at $180-200
-
Market share gains validated; 25% of AI accelerator market
-
Hold for $280-300 as recovery matures
-
Oracle (ORCL): 5%
Phase 3 Defensive Allocation (30%):
-
Bonds & Treasuries: 20%
-
Reduced but maintained for rebalancing flexibility
-
Lower yields (4-4.5%) as Fed cuts post-recession
-
Source of funds for final equity deployment
-
Cash: 10%
Phase 3 Success Metrics
If Scenario 3 Strategy Executed Perfectly:
Phase 1 (2025-2026):
-
Market falls 30%, your portfolio falls 10-15%
-
Outperformance: +15-20 percentage points
-
Short positions contribute +8-12% returns
Phase 2 (2027-2028):
-
Market recovers 20-25% from lows, you capture 50-60% (smaller equity allocation)
-
Portfolio up +10-15% while market still down 10-15% from 2025 peak
-
Cumulative outperformance: +20-25 percentage points
Phase 3 (2029-2030):
-
Market completes recovery, you participate in 60-70% of upside with increased allocation
-
Portfolio up +25-30% as quality names rally
-
Final cumulative outperformance: +30-40 percentage points
Total Return 2025-2030:
-
Your Portfolio: +55-75% (9-12% CAGR)
-
Buy-and-Hold AI Infrastructure: +15-25% (3-5% CAGR)
-
Outperformance: +40-50 percentage points
-
Most importantly: You avoided 40-50% drawdown and slept well
Phase 3 Risk: The Recovery You’re Not Positioned For
What if Phase 3 looks different than expected?
-
Recovery Concentrates in Names You Don’t Own:
-
New winners emerge (e.g., quantum computing, neuromorphic chips)
-
Your Phase 2 survivors (NVDA, MSFT) lag new technology
-
You preserved capital but miss next wave
-
Recovery Happens Faster Than Expected:
-
Market rebounds in 12 months (2026) instead of 24-36 months
-
You’re still 70% defensive when market is up 40%
-
Timing failure costs you 20-25% returns
-
No Clear Bottom Signal:
-
Market muddles sideways 2026-2028, never giving “all clear”
-
You stay defensive too long, miss gradual recovery
-
Death by waiting for perfect entry
Active Management Requirements for Scenario 3
Phase 1 (Defensive):
-
Daily: Monitor short positions, stop losses, volatility
-
Weekly: Market sentiment, positioning data, technical levels
-
Monthly: Thesis validation, trigger point assessment
-
Time Commitment: 10-12 hours/week (more than Scenario 2 due to short management)
Phase 2 (Selective Re-Entry):
-
Weekly: Entry points, valuation screens, sentiment analysis
-
Monthly: Dollar-cost averaging execution, position building
-
Time Commitment: 6-8 hours/week
Phase 3 (Aggressive Redeployment):
-
Weekly: Portfolio rebalancing, new opportunity identification
-
Monthly: Performance review, risk assessment
-
Time Commitment: 4-6 hours/week (back to normal active management)
Psychological Requirements
Scenario 3 is the most psychologically demanding strategy:
Phase 1: You watch the market potentially rally 20-30% while you’re defensive. Every day you question if you’re wrong. FOMO is intense.
Phase 2: You’re buying into a falling market. Catching falling knives is terrifying. Every entry looks wrong initially.
Phase 3: You’re buying back near old highs. You feel like you’ve wasted 3-4 years. Temptation to abandon strategy is huge.
Required Traits:
-
Extreme conviction in thesis
-
Willingness to look foolish for extended periods
-
Discipline to stick to predetermined trigger points
-
Emotional fortitude to manage shorts during rallies
-
Patience to wait months/years for thesis validation
Expected Outcome (2030)
Base Case (If Scenario 3 Thesis Correct):
-
Portfolio value: +55-75% (9-12% CAGR)
-
AI Infrastructure Buy-and-Hold: +15-25% (3-5% CAGR)
-
Absolute outperformance: +40-50 percentage points
-
Maximum drawdown: -15% (vs. -45% for buy-and-hold)
-
Sleep quality: Excellent throughout cycle
Success Defined: You preserved capital during a speculative bubble deflation, re-entered at attractive valuations, and participated meaningfully in the recovery. Most importantly, you avoided permanent capital loss and maintained optionality throughout.
Bear Case (If Wrong About Crash):
-
Portfolio value: +20-30% (4-5% CAGR)
-
AI Infrastructure Buy-and-Hold: +120-150% (17-20% CAGR)
-
Underperformance: -90 to -120 percentage points
-
Opportunity cost: Catastrophic
-
Emotional damage: Severe - watching from sidelines during generational wealth creation
What Success Looks Like
In 2030, you’re vindicated. Friends who stayed fully invested in AI suffered through 2026 crash, panic-sold at the bottom, and never recovered. You preserved capital, maintained discipline, re-entered methodically, and now have both capital gains and sleep-filled nights to show for it.
You didn’t get rich quick. But you didn’t get poor either. And now you’re positioned for the next cycle with capital, confidence, and a proven framework.
Suitability
Ideal for:
-
Investors with strong bear thesis conviction
-
Those who experienced prior bubble losses (2000, 2008, 2021) and learned lessons
-
Ability to withstand FOMO and social pressure
-
10-12 hours weekly for short position management
-
Emotional resilience to be wrong publicly for 12-24 months
Not suitable for:
-
Optimists about AI’s transformational potential
-
Inability to manage short positions (technical skill required)
-
FOMO susceptibility or herd mentality
-
Those who can’t tolerate looking “wrong” for extended periods
-
Anyone without conviction in mean reversion
