AI Data Center Infrastructure & Bubble Risk Simulator

Stress-test hyperscaler deployments under soaring GPU rack power densities (up to 120 kW/rack), 15-year utility power contracts, rapid 4-year hardware obsolescence, and sudden token demand shocks.

Hyperscale AI Financial & Physical Stress Matrix

Moderate Debt Stress
Total Initial CapEx
$16.0 B
$9.6B Debt / $6.4B Equity
Total Server Racks
5,882 Racks
Direct Liquid Cooling Req.
Annual Power Bill (PPA)
$361 M/yr
8,760 hrs @ 1.10 PUE
Net Stranded Asset Risk
-$5.4 B
Breakeven: Never (Default)
10-Year Free Cash Flow & Debt Paydown Bubble Shock Year marked
Rack Power Density vs. Cooling Envelope Extreme Liquid
Year Utilization Revenue ($M) Power PPA ($M) Opex + Maint ($M) Debt Service ($M) Net FCF ($M) Cumul. Balance ($M)
Vulnerable Hyperscale Trajectory: High leverage of 60% combined with a 2027 demand cliff will cause debt service to exceed post-crash HPC revenues, triggering balance sheet write-downs of $5.4B.

The Physical Paradox: 120 kW Racks

Traditional data centers run 5 to 10 kW per rack on air cooling. AI clusters packing Nvidia H100 and Blackwell servers generate 40 to 120 kW per rack. Standard chillers fail under this thermal load, requiring massive direct-to-chip liquid cooling loops and closed-circuit evaporative towers.

The Take-or-Pay Power Trap

Utilities demand 10-to-20 year take-or-pay power purchase agreements before expanding transmission or restarting nuclear units. If generative AI inference revenue slows, operators cannot simply turn off power without paying hefty contractual standby penalties.

Post-Crash Repurposing

Unlike empty fiber lines in the 2001 dot-com crash, depreciated GPUs can run molecular dynamics, weather simulation, and scientific rendering. However, rapid chip obsolescence means hardware loses 70% of peak resale value every 3.5 years.

AI Data Center Debt & Cooling Stress Dynamics

Read the explanation

Hyperscale AI facilities combine massive capital expenditure with high rack densities, scaling cooling needs and debt obligations simultaneously. During peak boom years, high compute utilization yields positive free cash flows, amortizing debt and covering fixed take-or-pay power contracts. Selecting the Bubble Shock preset in Year 3 slashes utilization to 35 percent. Unbending debt and take-or-pay electricity trigger a deep ten billion dollar stranded loss.

Enjoy this tool? Build your own with Super