The model, step by step
For a 30-day month, available hours are 720. Paid hours = 720 × paid utilization. Revenue = paid hours × hourly rate; subtract the platform fee only if that rate is before fees. Energy = 720 × [idle watts + utilization × (rented watts − idle watts)] ÷ 1,000. Operating surplus is receipts after fees minus electricity and your other monthly costs.
The default is an illustrative $0.20 hourly rate before a 10% fee, 30% utilization, 80 W idle, 350 W rented, $0.20/kWh electricity and $5 other costs. It produces $43.20 before fees, $4.32 of fees, 115.92 kWh of energy, $23.18 electricity and $10.70 operating surplus. These inputs do not describe a measured GPU model or marketplace average.
Why break-even utilization matters
Every rented hour earns after-fee revenue but also raises power draw above idle. The model solves for the utilization that covers idle power and other standing costs. If rented-hour receipts cannot cover the incremental power cost, or the required share exceeds 100%, it reports that the running costs cannot be recovered under those inputs.
Use evidence from your own host
Record paid GPU-hours, settled receipts, platform deductions, wall-meter energy and additional expenses over the same period. A manufacturer's GPU power rating excludes the rest of the computer and does not measure your host's energy use. Public asking prices cannot establish booked demand. The Vast.ai hosting documentation describes host responsibility for pricing, maintenance and availability during rental contracts; read your platform's current agreement before listing a machine.
Our GPU hosting budget guide shows a worked utilization comparison and a checklist for reconciling actual payouts. Financing, tax, equipment failure, resale, labor and storage or bandwidth revenues are excluded unless you account for them separately.