RTX 5090 from $0.09 per GPU-hour
32 GB of GDDR7 and the highest consumer bandwidth available. For single-GPU inference, image and video generation and small fine-tuning runs, nothing else in this catalogue gives so much throughput per dollar.
Billed per started minute. 128 GB of GPU memory on a 4× node. Persistent storage $0.08 per GB-month. No egress fees.
- GPU memory
- 32GBGDDR7
- Bandwidth
- 1,792GB/sSets token throughput on inference
- Node sizes
- 1× · 2× · 4×Same price per GPU at every size
- Reclaim rate
- 10–15%Spot instances reclaimed over 30 days
What you get with each RTX 5090 node.
| Node | GPU memory | vCPU | RAM | Local NVMe | Spot | On-demand | Action |
|---|---|---|---|---|---|---|---|
| 1× RTX 5090 | 32 GB | 8 | 64 GB | 500 GB | $0.09/h | $0.19/h | Configure |
| 2× RTX 5090 | 64 GB | 16 | 128 GB | 1 TB | $0.18/h | $0.38/h | Configure |
| 4× RTX 5090 | 128 GB | 32 | 256 GB | 2 TB | $0.36/h | $0.76/h | Configure |
GPU-to-GPU link: PCIe Gen5 x16. Local NVMe is scratch space wiped when the instance ends; keep anything you need on a persistent disk.
128 RTX 5090 GPUs free right now.
Straight from the capacity pool the console books against. Prices are identical in every region — pick the one closest to your data.
A region with zero free GPUs still accepts reserved capacity requests — we hold hardware for a term rather than sell what is already taken.
What fits in 32 GB — and what it costs per hour.
Weights need about 2 GB per billion parameters at FP16 / BF16, 1 GB per billion parameters at FP8, 0.5 GB per billion parameters at INT4. We keep 20% of the memory free for activations, the KV cache and the CUDA context, then take the smallest node that still fits.
| Open model | Parameters | FP16 / BF16node · spot price | FP8node · spot price | INT4node · spot price |
|---|---|---|---|---|
| Mistral 7BAssistants, RAG, classification | 7.2B | 1× $0.09/h | 1× $0.09/h | 1× $0.09/h |
| Llama 3.1 8BAssistants, agents, fine-tuning | 8B | 1× $0.09/h | 1× $0.09/h | 1× $0.09/h |
| Gemma 2 27BHigher-quality assistants | 27B | 4× $0.36/h | 2× $0.18/h | 1× $0.09/h |
| Qwen2.5 32BCode and reasoning | 32B | 4× $0.36/h | 2× $0.18/h | 1× $0.09/h |
| Mixtral 8x7BMixture of experts, high throughput | 46.7B | 4× $0.36/h | 2× $0.18/h | 1× $0.09/h |
| Llama 3.3 70BThe common production baseline | 70B | over 4× | 4× $0.36/h | 2× $0.18/h |
| Qwen2.5 72BMultilingual, long context | 72B | over 4× | 4× $0.36/h | 2× $0.18/h |
| Mixtral 8x22BMixture of experts, large capacity | 141B | over 4× | over 4× | 4× $0.36/h |
| Llama 3.1 405BLargest widely used open model | 405B | over 4× | over 4× | over 4× |
Real usage depends on context length, batch size and the serving engine: a long context can add tens of gigabytes of KV cache. Treat the table as the floor, not the ceiling — and if a job is close to the limit, take the next node size or quantise one step further.
What a dollar buys on an RTX 5090.
Ranked 4 of 14 consumer models in this catalogue on cost per gigabyte of GPU memory.
GPU memory you get for $1.00 of spot time, per hour.
Bandwidth decides token throughput far more often than raw FLOPS.
Spot price divided by published FP32 throughput.
One GPU on spot, billed per minute, storage excluded.
Against the rest of the market
34 providers · September 2026- Median on-demand elsewhere
- $0.62Our on-demand is 69% below it
- Cheapest on-demand seen
- $0.25Our on-demand $0.19 is 24% below it
- Cheapest spot seen
- $0.12Our spot $0.09 is 25% below it
- 24 hours on one GPU
- $2.16 on spot$4.56 on-demand · $6.00 at the cheapest rate found elsewhere
Index built from published prices for the same GPU across 137 providers (public price index (getdeploying.com)), September 2026. How the guarantee is enforced.
What people run on an RTX 5090.
Fine-tuning small models
LoRA on 7B–13B models, and full fine-tuning below 3B.
Image and video generation
Diffusion pipelines run entirely in memory at this capacity.
Anything that can checkpoint
On the spot tier this model is 53% below its own on-demand price, with a 2-minute notice before a reclaim.
NVIDIA RTX 5090, as published by NVIDIA.
Taken from the vendor datasheet. Where a figure is not published, the row is absent rather than estimated.
- Vendor
- NVIDIA
- Architecture
- Blackwell (2025)
- Segment
- Consumer
- Memory
- 32 GB GDDR7
- Memory bandwidth
- 1,792 GB/s
- FP32
- 105 TFLOPS
- CUDA cores
- 21,760
- Board power
- 575 W
- Form factor
- PCIe
- Interconnect
- PCIe Gen5 x16
- Multi-instance (MIG)
- Not available
- 8-bit float (FP8)
- Supported by the architecture
Source: images.nvidia.com · www.nvidia.com · verified against the vendor document
Software that runs on it
8 templatesPre-built environments, pulled on the node before you land on it. Or bring any OCI image from a public or private registry.
On the spot tier
10–15% reclaimed · 30 d- 2 minutes of notice on the metadata endpoint, a webhook and the console
- The instance is stopped, not deleted — disk and IP stay attached
- Auto-relaunch on the next free RTX 5090, or switch the same disk to on-demand
- You save $2.40 a day per GPU against on-demand
If the RTX 5090 is not the right fit.
Same memory or more, 33% cheaper on spot
A100 40GB 40 GB HBM2 · Ampere $0.09/h spotNext step up in memory: 40 GB instead of 32 GB
RTX 5080 16 GB GDDR7 · Blackwell $0.06/h spotSame Blackwell generation
RTX 5070 Ti 16 GB GDDR7 · Blackwell $0.06/h spotSame Blackwell generation
Or put two of them side by side: RTX 5090 vs V100 · RTX 5090 vs A100 40GB · RTX 5090 vs RTX 5080
RTX 5090 — the questions that come up
How much does an RTX 5090 cost per hour?
Spot is $0.09 per GPU-hour and on-demand is $0.19, both billed per minute — every started minute costs the hourly price divided by 60, so $0.0015 on spot. Reserved capacity is $0.09 per GPU-hour on a monthly commitment. Storage and egress are not included in that rate: persistent disks are $0.08 per GB-month and there are no egress fees.
How many GPUs can I put in one instance?
Node sizes are 1×, 2×, 4× — up to 4 RTX 5090 GPUs in a single virtual machine, with CPU, RAM and local NVMe scaled with the GPU count. The price per GPU is identical at every node size.
Can it run a 70B model?
Yes. At FP8, a 70B model needs about 70 GB for the weights, so it fits on 4× RTX 5090 ($0.36 per hour on spot) with roughly 20% of the memory left for activations and the KV cache.
Which regions have it?
Available now in US West (Oregon), EU Central (Frankfurt). Prices are identical in every region; pick the one closest to your data.
How often is a spot RTX 5090 reclaimed?
Over the trailing 30 days, 10–15% of spot instances on this model were reclaimed. You get a 2-minute notice on the metadata endpoint, through a webhook and in the console; the instance is then stopped, never deleted, and the persistent disk stays attached.
Anything else about this model? Ask an engineer — the same people run the nodes.
An RTX 5090 for $0.09 an hour, running in under 60 seconds.
Pay as you go — no contracts, no minimum commitment. Add credit, launch, stop whenever you want.
Billed per minute from the moment the instance is reachable. Minimum credit $40, no subscription. 128 RTX 5090 GPUs available right now.