Data Center/AI Investment Viability Model

Investment & Costs

Capacity source

Up-front investment ($) GPU % Construction % Other % (power, network, land) Cost per GPU ($)

Training / R&D share of capacity (%)

GPUs / GPU-equivalents:

Fixed operating cost ($ / month)

Energy & electricity market

Power draw per GPU (kW) PUE (cooling overhead) Electricity price ($ / kWh) Base price growth (% / yr) Regional grid capacity (MW) Project price pressure (%/yr per 1% load share)

Energy cost: / month (month 1; varies with load)

Financing & depreciation

Debt-financed portion (%) Interest rate (% / yr) Loan type Loan term (years) Discount rate for NPV (% / yr) GPU useful life (years) Construction useful life (years) Other useful life (years) Opex inflation (% / yr) Committed spend growth (% / yr, rented mode)

Show electricity cost curve

Usage Projections

The Cloud↔Local slider sets what share of each task earns this data center revenue — local tasks run on users' devices. Raw volumes, split growth rates, and token overrides live under each card's Advanced settings.

Advanced settings — model pricing mix, serving capacity, horizon

Model served	Share %	In $/MTok	Out $/MTok	Decay % / yr

Model-mix shares auto-normalize; hover a model name for its price source. Blended price now: / MTok.

Max utilization (%) Throughput per GPU (tokens/s) View Horizon (years) In:out token ratio

Serving capacity: tokens / month at max utilization