Best GPU servers for self-hosted AI
Which cloud GPU server to rent for running your own AI models, from a single RTX 4090 to H100s, with hourly prices compared.
Updated
This guide is being written. Below is the outline and our current price data; test results will be added once our testing is complete.
Quick comparison
| Provider | From | Starting plan | Free tier | Our rating | Link |
|---|---|---|---|---|---|
| RunPodChecked | $0.34/hr | RTX 4090 24GB (Community Cloud) | No free credits; small referral bonus after a $10 top-up | Not rated | VisitOfficial site |
| Vast.aiChecked | $0.49/hr | RTX 4090 24GB | No free credits | Not rated | VisitOfficial site |
| LambdaChecked | $2.29/hr | GH200 96GB (1x instance)64 vCPU · 432 GB RAM | No | Not rated | VisitOfficial site |
| HyperstackChecked | $1.35/hr | A100 80GB (PCIe)A100 PCIe · 28 vCPU · 120 GB RAM | No | Not rated | VisitOfficial site |
| DigitalOcean GPU DropletsChecked | $0.76/hr | RTX 4000 Ada 20GB8 vCPU · 32 GB RAM | New-account credit; GPU use not confirmed | Not rated | VisitOfficial site |
| Vultr Cloud GPUChecked | $1.67/hr | L40S 48GB (1 GPU)16 vCPU · 180 GB RAM | New-customer promo credit; GPU use not confirmed | Not rated | VisitOfficial site |
| TensorDockChecked (partly unverified) | $0.35/hr | RTX 4090 24GB | No ($5 minimum deposit) | Not rated | VisitOfficial site |
The short answer
Our pick for most people, the cheapest option that works, and the pick for serious training: one line each, linked to the comparison below.
How much GPU do you need?
A table that maps model size (7B, 13B, 32B, 70B and up) and precision (16-bit, 8-bit, 4-bit) to the VRAM needed, and the cheapest GPU that fits.
GPU by GPU
- RTX 4090 / 5090 (24–32 GB): best value for inference of small and mid-size models.
- L40S (48 GB): a middle ground for bigger models.
- A100 80 GB: the long-time workhorse for training and large models.
- H100 / H200: fastest, for big training runs and high-throughput serving.
Provider by provider
What each GPU costs per hour at each provider, plus what the price leaves out: storage, network egress, minimum billing.
Dedicated server vs cloud GPU
When buying or leasing a dedicated GPU server beats renting by the hour.
Our tests
Tokens per second for the same open model on each provider. [Results to be added after testing.]