AI compute planning, without fake live prices

Know the bill before the run.

Compare GPU-hour costs, VRAM fit, utilization, storage and egress. Every rate is editable, so you can paste in the current quote from any provider.

Build a workload

GPU / memory baseline

Required total VRAM (GB)

GPU count

Runtime per day (hours)

Billing days

Expected utilization (%)

Storage cost for period (USD)

Egress / extras (USD)

Provider options — name, hourly USD per GPU, VRAM per GPU

What this calculates

Compute cost equals hourly GPU rate × GPU count × scheduled hours × utilization. Storage and egress are added as user-supplied period costs.

VRAM fit check

An option is flagged when GPU count × VRAM per GPU is below the stated requirement. Real workloads may also need model, KV-cache and framework overhead.

What it cannot predict

Spot interruptions, regional availability, data-transfer tiers, taxes, reserved discounts, startup time and vendor-specific billing increments must be checked separately.