NVIDIA RTX 5090 Pricing
Eli Mernit
An NVIDIA RTX 5090 costs $0.40 to $0.72 per GPU-hour to rent on demand from the providers below, against a market median of $0.64 across 15 providers. Beam's RTX 5090 machine is $0.72 an hour, the top of that range, with vCPU, RAM, and NVMe included. Beam's serverless 5090 is $1.09 an hour for the GPU, plus CPU and RAM, under a committed-spend agreement.
Key takeaways
- Vast.ai (from $0.40) and Salad ($0.50, High priority) are both cheaper than Beam's $0.72 on-demand price.
- Spheron ($0.68) and RunPod Community Cloud ($0.69) also list lower rates, but neither has stock.
- Beam's machine is $0.08 above the market median.
RTX 5090 cloud prices per hour, by provider
| Provider | SKU | On-demand $/GPU-hr | Notes |
|---|---|---|---|
| Vast.ai | RTX 5090 (marketplace) | $0.40 | Cheapest rentable offer; median offer $0.63; spot from $0.10 per getdeploying; marketplace data from Vast.ai |
| Salad | RTX 5090 (32 GB), High priority | $0.50 | Consumer-PC network; vCPU and RAM included; interruptible tiers from $0.25 |
| Spheron | RTX 5090 (1x, 8 vCPU, 24 GB RAM) | $0.68 | No capacity listed; last listed rate; per-minute billing, 20-minute minimum |
| RunPod | RTX 5090 (Community Cloud) | $0.69 | Out of stock per getdeploying; Secure Cloud $0.99, in stock |
| Beam | RTX 5090 PCIe | $0.72 | On-demand machine; vCPU, RAM, and NVMe included; serverless GPU $1.09/hr with a committed-spend agreement, plus CPU and RAM |
| Market median | — | $0.64 | Across 15 providers with an on-demand price (getdeploying) |
How these prices were collected
Each figure is an on-demand, per-GPU list price for a single-GPU shape where one is offered, with spot and interruptible rates kept to the notes. Marketplace data from Vast.ai.
- Vast.ai's GPU pricing page has the floor and median offers.
- Salad's pricing page has each priority tier.
- Spheron's RTX 5090 pricing page shows the rate and capacity.
- RunPod's pricing page has both cloud tiers.
- Beam's pricing page has the on-demand and serverless rates.
- getdeploying's RTX 5090 tracker supplies the median.
RTX 5090 vs RTX 4090 vs L40S: where the 5090 fits
The RTX 5090 is a Blackwell card with 32 GB of memory, between the 24 GB RTX 4090 and the 48 GB L40S. Market medians run $0.47 for the 4090, $0.64 for the 5090, and $1.54 for the L40S, so moving up from a 4090 costs $0.17 an hour for 8 GB more memory and FP4 support.
On Beam, the 5090 machine is $0.72 and the L40S is $0.76, so $0.04 more buys 16 GB more memory. At market medians, the L40S costs 2.4 times as much. The 5090's edge on Beam is serverless, which the L40S doesn't offer.
Where to rent an RTX 5090 for less
Yes. Vast.ai and Salad both beat Beam's $0.72 on-demand price. Vast.ai's marketplace data shows offers from $0.40, and even its $0.63 median is below Beam. Salad's High priority tier is $0.50 with vCPU and RAM included.
The trade-offs: every Vast.ai offer is a different host's machine, and Salad's consumer PCs can disconnect even at High priority. Spheron and RunPod Community Cloud also undercut Beam but show no stock. Beam's $0.72 buys a dedicated machine with NVMe included.
What this means for your workload
The 5090 suits inference on models that outgrow 24 GB but fit in 32 GB, including FP4-quantized ones. For steady traffic, Salad and Vast.ai undercut Beam's machine if their terms fit your job. Beam's serverless 5090 suits bursty traffic, but its GPU rate plus CPU and RAM charges mean it only wins when the GPU would otherwise sit idle.
FAQ
How much does an RTX 5090 cost per hour?
It costs $0.40 to $0.72 per GPU-hour on demand across the providers above, against a $0.64 market median. Beam's machine is $0.72 an hour; serverless is $1.09 an hour for the GPU with committed spend.
What's the cheapest way to rent an RTX 5090?
Vast.ai's marketplace is cheapest, with offers from $0.40 and a $0.63 median. Salad's High priority tier is $0.50, and its interruptible tiers start at $0.25 if your job can be preempted.
RTX 5090 vs L40S: which should I rent?
Rent the 5090 when 32 GB is enough; its median is $0.64 against $1.54 for the L40S. On Beam the gap is only $0.04 an hour, so price the L40S there if you need 48 GB.
