AI Supercycle Stack Economics
Stanford Engineering Value Chain & Energy Bottleneck Model
Export CSV
Export JSON
Reset State
Scenario Presets:
Default Baseline
Energy Grid Bottleneck
GPU Oversupply / Hyperscale
Model Commodity Pressure
App Layer Monopolization
Capital & Physical Inputs
Total Capital Allocated ($B)
$150B
Power Grid Capacity (GW)
12.5 GW
Data Center PUE Efficiency
1.25
Deployed GPU Clusters (k Units)
250k
Model Scaling (FLOPs 10²⁶)
5.0
App Layer End Demand ($B)
$45B
6-Layer Stack Flow & Value Capture
D3 Dynamic Flow
1. Energy Grid
2. Data Centers
3. Hardware/GPUs
4. Models
5. Middleware
6. Applications
Primary System Constraints
Primary Active Bottleneck
Energy & Power Distribution
Grid interconnection limits total cluster compute capabilities below invested hardware capacity.
Power Utilization
94.2%
Token Throughput
4.20T/s
Hardware ROIC
18.5%
App Layer ROIC
24.1%
Layer Returns Breakdown
Layer
CapEx ($B)
Margin %
ROIC %
Canonical Model State Proof
Focus Proof Surface
Primary Bottleneck
Energy & Power Distribution
Energy Utilization
94.2%
Token Throughput (TPS)
4.2E+12
Hardware ROIC
18.5%
App Layer ROIC
24.1%
Model Layer Margin
12.4%