cheapcompute
GPU capacity, on demand
Compute without the cloud premium

Bare-bones marketplace for cheap compute

Grab unused GPU capacity for batch and training jobs. Clear windows, predictable rates, and a human in the loop when you need one.

H100$1.90/hr · us-west · tonightA100$1.25/hr · eu-central · rolling blocks4090$0.65/hr · us-east · next 48hSpin-up< 30 min · human assisted H100$1.90/hr · us-west · tonightA100$1.25/hr · eu-central · rolling blocks4090$0.65/hr · us-east · next 48hSpin-up< 30 min · human assisted
Live GPUs
268
H100 / A100 / 4090
Typical rate
$0.65–$1.90/hr
Region-dependent
Spin-up
< 30 min
Human-assisted scheduling
GPU-first
A100s, H100s, 4090s—and growing. CPU capacity available for supporting work.
Fast spin-up
Tell us your job and we reserve a window. No cluster babysitting required.
Predictable billing
Clear hourly rates and time blocks so batch runs stay on budget.
How it works
A simple path from request to run. We keep humans in the loop when it matters.
Step 1
Tell us your job
Model, budget ceiling, duration, and timing.
Step 2
We match capacity
Find the right GPU/region and reserve a window.
Step 3
Ship your workload
Run via your stack; we handle the handoff.
Sample slots
Current availability and ballpark rates.
H100
80GB · us-west · low intra-region latency
$1.90/hr
us-westTonight 9p–3a PT
A100
40GB · eu-central · balanced throughput
$1.25/hr
eu-centralRolling 12h blocks
4090
24GB · us-east · flexible windows
$0.65/hr
us-eastNext 48h
Why teams use cheapcompute
Keep experiments moving without renting entire clusters.
Spin up training windows overnight; we schedule the slot.
Mix GPU types to fit budget and job shape—no cluster babysitting.
We vet hosts, handle ops, and keep humans in the loop.
Direct line for support—no ticket maze.
FAQ
Quick answers; reach out for anything deeper.