Pricing
Pay for the seconds you use
One price per GPU-hour, billed by the second. Storage and networking between our regions are included. Commit for longer and save up to 35%.
GPU
Per GPU-hour
Spark 24
$0.69
Spark 48
$1.19
Forge 80
Popular
$2.29
Forge 141
$3.09
Anvil 192
New
$4.49
Spot clusters are up to 60% off and can be paused with two minutes’ notice.
Calculator
What will my cluster cost?
Self-serve
Pay as you go
$0
platform fee
Any GPU, any size
Billed by the second
Community support
Most teams
Reserved
−35%
with a year
Guaranteed capacity from 64 GPUs
A named support engineer
Private networking included
Large scale
Dedicated
Custom
pricing
A whole hall, built to your spec
On-site engineers
Bring your own hardware
Questions
Pricing, plainly
Do you charge for storage or data transfer?
The first 10 TB of storage per cluster and all transfer between our regions are included. Egress to the internet is $0.02 per GB after the first 5 TB a month.
What happens if a GPU fails mid-run?
We swap the node and restart your job from the last checkpoint automatically. You aren’t billed for the time the node was down.
Can I mix commitments?
Yes. Many teams reserve a base cluster for a year and burst on demand or spot above it.
Is there a free tier?
New accounts get 100 GPU-hours on Spark 24 or 25 on Forge 80, whichever you use first.
Start in minutes
Your first 100 GPU-hours are on us
Sign up, add a card and launch a cluster before your coffee’s cold. No sales call, no minimum spend.