AI Infrastructure CapEx & Safeguard Pace Simulator

HEARING EXCERPT // TESTIMONY

“AI companies can stop whenever they want to.” — Treasury Secretary Scott Bessent

Ranking Member Maxine Waters challenged the thesis: once hyperscale capital & gigawatt grid commitments are contracted, is a voluntary pause economically or contractually possible?

Scenario Presets

Simulated Levers

Cluster Power Draw 12.5 GW

1 GW (Metro) 35 GW (Multi-state Grid)

Hyperscale CapEx Burn $45.0 B/yr

$10B $150B

Voluntary Pause Readiness 35 %

Contract Lock (5%) Full Discretion (95%)

Congressional Oversight Pressure 78 / 100

Laissez-Faire (10) Subpoena Climax (100)

BESSENT THESIS CHECK: When CapEx burn exceeds $30B/yr and long-term power purchase agreements (PPAs) exceed 10 GW, contractual momentum drops voluntary stoppage feasibility below 30%, making "they can stop whenever" an empirical illusion.

System Status

High Capital Momentum / Grid Constrained

Voluntary Feasibility

28.5 %

Sunken CapEx Locked

$ 142.5 B

Congressional Exposure

84 / 100

Causal Force Field Drag Nodes to Stress-Test

Compute/CapEx Grid Bottleneck Oversight Force

Momentum Inertia: High

Grid Lock Risk: Severe

Bessent Feasibility Ratio: 0.29x

1. The "Can Stop Anytime" Fallacy

Hyperscale AI companies sign 15-to-20 year Power Purchase Agreements (PPAs) and custom substation capital commitments. Halting computation does not eliminate financial debt servicing, transformer reservation costs, or hardware depreciation write-downs.

2. Maxine Waters Oversight Vectors

Congressional testimony probes systemic financial contagion: if AI clusters fail to generate sufficient operational software returns to service their grid and debt commitments, does the systemic risk migrate to regional utilities and commercial debt markets?

Simulated Levers

1. The "Can Stop Anytime" Fallacy

2. Maxine Waters Oversight Vectors

Executive Brief Export