Cloud GPUs
GPU compute, billed
straight from your wallet.
Pick a machine and region, connect your wallet, and launch. Usage is metered by the hour and billing is handled through Pulze, our billing partner. An initial reservation is applied to your metered usage.
How billing works
The meter runs. The agent settles. You get the receipt.
Every hour, an autonomous agent closes your usage windows and sends the metered amount to Pulze, our billing partner. You receive an itemised receipt for what actually ran.
- Meter
Hourly windows, hash-keyed
- Verify funds
Pulze runway check
- Charge mandate
Capped, on-chain
- Receipt
Itemised, emailed
One H100 SXM at $2.49/hour. The agent settles at $25, never above the cap the customer signed, and the receipt lands before the funds clear.
Pick the silicon. See what it costs.
Six accelerators, published rates, billed by the metered hour. Select one to see the board and what a run actually costs.
Billed per metered hour against your mandate. No egress fees, no minimums, no reserved-instance contract to sign.
- VRAM
- 80 GB HBM3
- System RAM
- 125 GB
- vCPU
- 20
$5.58 is charged at launch as a 2-hour reservation. It proves the mandate works and comes straight off your first metered hours.