# Virya Compute Risk Explorer

> Models what an AI data center earns or loses per GPU-hour (the compute spread) in 86 economies, at site sizes of 1 MW, 100 MW, and 1 GW. Modeled scenarios, not investment advice.

## What it computes
- Compute spread: GPU rent revenue per installed GPU-hour, idle hours included, minus electricity, GPU payback, building (land, building, and grid connection), staff and admin, and a country risk charge in two layers (customer credit, political and transfer risk).
- GPU rent is the price per GPU-hour a customer pays; land, building, and grid connection are costs in the building line, not GPU rent.
- Break-even utilization, GPU rent, and power price; stress tests at 20% lower GPU rent and one year less GPU life (no resale value at the 3-year floor).
- Summary PDF only: a cost table, illustrative protection pricing (four floors, outage cover, a power hedge; not quotes) in the combined downside, and a yearly cash table.

## Site sizes (grid power)
GPUs = grid / PUE / server kW per GPU. At PUE 1.3: 1 MW, 603 H100 (430 B200, 947 A100); 100 MW, 60,332; 1 GW, 603,318 on 769 MW of IT load

## Inputs
GPU (A100, H100, H200, B200); workload (training, fine-tuning, inference); list GPU rent, GPU rent decline, utilization, GPU rent multiplier, contracted share; months until first revenue; GPU capex, GPU life, resale value; electricity price, PUE, building cost, labor cost index, Δ cost of capital; GPU rent volatility, utilization spread, insurer loading.

## Sources
Virya Electric Grid Index, World Bank, ILOSTAT, Eurostat, IMF, Damodaran, national regulators and utilities, European Central Bank (ECB) reference rates, and ExchangeRate-API. Exchange rates refresh twice a day (ECB, else ExchangeRate-API); other inputs are reviewed periodically.

## Links
- [Explorer](https://spreads.anshulsaxena.com/)
- Methodology details on request: contact@anshulsaxena.com

## Contact
Modeled scenarios, not investment advice. Feedback or questions: contact@anshulsaxena.com