Deploy NVIDIA H100 and A100 instances in seconds. True per-second billing, full root access, and instant Docker deployment — no minimums, no lock-in.
Platform
Connect over encrypted SSH with key-based authentication and isolated networking.
Launch any image or pull from our optimized template registry in under 30 seconds.
Complete administrative control — install drivers, tune the kernel, anything you need.
Attach durable NVMe volumes that survive restarts so datasets and checkpoints stay put.
Automate provisioning, scaling, and teardown with a clean REST API and first-class CLI.
Stop paying for idle time. Metering stops the moment your pod is terminated.
Pricing
| GPU model | VRAM | TFLOPS (FP16) | On-demand | |
|---|---|---|---|---|
| NVIDIA H100 SXM | 80 GB | 1,979 | $2.49/hr | Deploy → |
| NVIDIA A100 | 80 GB | 624 | $1.29/hr | Deploy → |
| NVIDIA RTX 4090 | 24 GB | 330 | $0.58/hr | Deploy → |
| NVIDIA RTX 3090 | 24 GB | 142 | $0.31/hr | Deploy → |
Interruptible (spot) instances are available at up to 60% lower cost. See full pricing.
Join thousands of ML engineers training and serving models on Poidex.