To The Moon

1. Arbitraging the Interconnection Queue

In 2026, the binding constraint for AI expansion is no longer GPUs—it’s grid interconnection queues.

Hyperscalers are effectively renting electricity through capital timing. Securing baseload rights ahead of peers creates an arbitrage on interconnection bottlenecks.

North Virginia and Texas grid interconnection queues, showing 3–5 year delays

2. 25GW Non-Discretionary Redistribution of Margin

Assume the AI sector adds 25GW of load:

• Annual consumption ≈ 219 million MWh

• PPA premium ≈ $60/MWh

Non-discretionary redistribution of margin ≈ $13B/year

This is a forced transfer of profits from hyperscalers to operators with existing nuclear assets.

[Insert Image 2: Bar chart showing PPA-induced profit redistribution across cloud giants]

3. Scarce Assets: Nuclear Operators in Focus

Commercial nuclear capacity in 2026 is concentrated:

• $CEG (Constellation Energy) – 100% nuclear, the “baseload king”

• $VST (Vistra) – mixed nuclear and gas, operational leverage to spark spreads

Scarcity rents are now the primary value driver, not AI growth.

4. CEG vs. VST: Structural Differentiation

$Constellation Energy Corp(CEG)$ : Stability through Baseload

• Pure nuclear fleet

• Predictable cash flow

• Secures long-term PPA pricing

$Vistra Energy Corp.(VST)$ : Operational Leverage to Spark Spreads

• Gas units allow margin capture in peak periods

• High elasticity of cash flow during scarcity events

CEG wins in stability, VST wins in price-flexibility leverage.

5. Contractual Rigidity Risk

Long-term PPAs represent a long-term liability overhang:

• Microsoft may lock in $150/MWh for 20 years

• 5 years later, cheaper power or AI efficiency gains could render these contracts structurally rigid

Track financial statements for Unconditional Purchase Obligations disclosures.

Californians urged to cut power use during extreme heat

6. Distribution Constraints: Copper and Transformers

Transmitting 5GW requires massive infrastructure:

• $GEV (GE Vernova) – grid and transformer lead times

• $ETN (Eaton) – power distribution systems

• $VRT (Vertiv) – internal data center power management

Vertical integration ensures priority delivery → justifies premium valuations.

7. Efficiency Hedge: Vertiv

• Reducing PUE by 0.01 can save tens of millions per data center

• Critical to offset rising electricity costs

8. Incremental AI-Revenue per GWh

• Tracks capital efficiency relative to electricity costs

• If power cost growth > incremental AI revenue per GWh → ROIC compression → valuation risk

Futuristic AI Data Center Power Demand

9. Regulatory Variable

• FERC reforms may shorten interconnection queues

• Reduced bottleneck → nuclear scarcity rents could compress

10. Minimalist Trading Checklist

1. Go long “time scarcity premium”: $CEG, $VST (add if nuclear restart progress exceeds expectations)

2. Go long “delivery resilience”: $GEV, $ETN (add if transformer lead times extend further)

3. Monitor risk threshold: $MSFT, $AMZN PPA averages

• If average exceeds $160/MWh and incremental AI-revenue growth slows → reduce hyperscaler exposure

11. Summary Verdict

In 2026, the equity risk premium for AI has migrated. We are transitioning from a “GPU Scarcity” regime to a “Grid Scarcity” regime.

• Longs: Nuclear IPPs ($CEG, $VST) for scarcity rents; Infrastructure incumbents ($GEV, $ETN) for backlog visibility

• Shorts/Cautious: Over-leveraged second-tier clouds and hyperscalers with unhedged power exposure

Bottom Line: You are no longer just buying a software company; you are buying a 20-year commitment to some of the most expensive utility bills in human history.