GPUs from $0.02/hr
10,000+ GPUs at the lowest market prices
Hyperscalers and other clouds have big margins and expensive high-end GPUs, leaving AI companies with a huge bill. SaladCloud offers instant, on-demand access to 10k+ consumer GPUs at the lowest prices in the market. Savor a transparent, predictable, and far more affordable pricing for AI/ML inference at scale.
GPU Pricing Calculator
Know exactly how much you will pay. Get exactly the resources you need.
All our instances are fully customizable for multi-GPUs, RAM & vCPUs.
Frequently Asked Questions
Simple & Transparent Pricing. No Surprises. Lowest in Market.
You only pay for the time that the hardware is available to be used by your application, you will never pay for Cold Boot time.
If you are looking to deploy 50+ GPU instances, we offer volume discounts. Please fill out the Demo Form to discuss discounted pricing.
We utilize unused, latent compute resources from gamers worldwide to power our network - think Airbnb for GPUs. The competitive marketplace helps keeps our prices low.
We do not charge for the initialization process (selecting hardware, downloading & loading container, etc). Once your container is running, charges start accruing.
GPU Instances
SaladCloud’s Nvidia GPU Instances provide additional computing power to accelerate your complex workloads.
All instances are fully customizable to your exact GPU, vCPU, and RAM requirements.
General Purpose Instances
General purpose instances are suitable for a wide variety of production workloads.
All instances are fully customizable to your exact vCPU and RAM requirmeents.
CPU-Optimized Instances
CPU-Optimized Instances provide a 2:1 ratio of memory to CPU. This configuration is optimal for applications that demand fast, consistent performance from dedicated vCPUs, but are not so memory-intensive that they require the additional RAM provided by General Purpose Instances.
All instances are fully customizable to your exact vCPU and RAM requirements.
Memory-Optimized Instances
Memory-Optimized Instances provide additional memory for each vCPU to help you avoid out-of-memory errors or excessive swaps to disk.
All instances are fully customizable to your exact vCPU and RAM requirements.