The whole card.
Only yours.
A Dedicated GPU is a full NVIDIA card assigned to your instance and nothing else, for as long as it runs. Steady throughput for training runs, benchmarks and latency-critical serving, from $0.39/hr.
A GPU you can
measure against.
Root access, the full card, and nothing else scheduled on it.
Single tenant
No one else's kernels share your card. Run-to-run variance comes from your code, not your neighbours.
Root and your stack
SSH, the web terminal or Jupyter. Install anything, or launch with a preset for vLLM, PyTorch or VS Code.
Hourly, no commitment
Billed by the hour from your wallet. Terminate and charges stop.
Dedicated or Dynamic?
Both, per job.
Same account, same wallet, same hardware.
| Dedicated | Dynamic | |
|---|---|---|
| The card | Yours alone | Pooled with isolated tenants |
| Price (B200) | $7.99/hr | $4.25/hr |
| Throughput | Steady | Varies with neighbours' bursts |
| GPU memory | All of it | The product's full allocation, enforced |
| Best for | Training, benchmarks, strict latency | Inference, fine-tuning, dev, agents |
| Monthly plan | 2× B200, $11,665.40/mo | RTX PRO 6000, $299/mo |
Bare-metal 8-GPU servers have no virtualization layer at all, with NVLink between GPUs.
Every GPU.
One list.
Dedicated is available on B200 (1, 2, 3 or 8 GPUs) and RTX 4090.
| GPU | Memory | Dynamic | Dedicated | Monthly | |
|---|---|---|---|---|---|
B200New NVIDIA · Blackwell | 180 GB HBM3e | $4.25/hr−47% | $7.99/hr | $11,665.40/mo for 2× | Details |
| NVIDIA · Blackwell | 96 GB GDDR7 | $0.76/hr | — | $299/mo | Details |
| NVIDIA · Ampere | 80 GB HBM2e | $0.89/hr | — | — | Details |
| NVIDIA · Ada Lovelace | 24 GB GDDR6X | — | $0.39/hr | — | Details |
Your next run
starts at $0.39.
Sign up, top up, launch. You'll be on a GPU in minutes.