Scenario definitions (24-month horizon)
1. No Impact AI hype unwinds. Hyperscaler capex normalizes to pre-2023 trend (~10% growth). Frontier model ROI questioned. GenAI adoption stalls. Beneficiary multiples re-rate to long-run averages.
2. Modest Impact AI is real but slower than bulls expect. Hyperscaler capex grows 15-20%/yr. Enterprise adoption is gradual (3-5% of seats by 2027). Compute scarcity persists but eases.
3. Strong Impact AI drives broad enterprise productivity. Capex grows 25-35%/yr. Frontier compute remains supply-constrained. Power becomes the binding constraint. Software margins expand as AI replaces seats.
4. Massive Impact AGI-adjacent productivity shock. Capex doubles. Power and chips both constrained for years. AI-native software displaces incumbents. Sovereign + private demand explodes. Datacenter buildout multi-trillion.