Skip to content
CloudTweaks

GPU Cloud

14 providers in this category.

GPU cloud providers rent accelerated compute (H100-class and similar) by the hour or second for training and inference. They differ in price, billing granularity, spot availability, and whether they offer serverless or bare-metal access. GPU rates move quickly, so compare a representative $/GPU-hour, per-second billing, spot, and bare-metal options against live pricing.

Companies

Neutral ordering. Not a recommendation.

14 results

CoreWeave

www.coreweave.com

CoreWeave is a cloud provider specializing in large-scale GPU compute for AI training and inference, built Kubernetes-native with bare-metal access.

$$$ Premiumusage-based pricing
Strengths
Maturity, operating 9 yearsFlexibility, Bare metal +1

Crusoe

www.crusoe.ai

Crusoe is an energy-first AI cloud that builds and operates its own datacenters, offering GPU compute on-demand and under contract with a vertically integrated power strategy.

$$ Midusage-based pricing
Strengths
Maturity, operating 8 yearsValue, Mid tier

DataCrunch

datacrunch.io

DataCrunch is a Finnish GPU cloud providing on-demand and reserved NVIDIA GPU instances, including H100 and newer generations, from Nordic data centers running on renewable energy.

$ Budgetusage-based pricing
Strengths
Maturity, operating 6 yearsValue, Budget tier

FluidStack

www.fluidstack.io

FluidStack provisions large dedicated GPU clusters for AI training, aggregating data-center capacity to deliver thousands of interconnected H100/H200-class GPUs to a single customer.

$$ Midusage-based pricing
Strengths
Maturity, operating 9 yearsValue, Mid tier

Genesis Cloud

www.genesiscloud.com

Genesis Cloud is a European GPU cloud focused on sustainable, renewable-powered AI compute, offering current-generation NVIDIA GPUs for training and inference.

$$ Midusage-based pricing
Strengths
Maturity, operating 8 yearsValue, Mid tier

Hyperstack

www.hyperstack.cloud

Hyperstack, operated by NexGen Cloud, is a European GPU cloud offering on-demand H100s and other NVIDIA GPUs from renewable-powered data centers.

$ Budgetusage-based pricing
Strengths
Price, Low hourly costMaturity, operating 6 years

Lambda

lambda.ai

Lambda is a cloud provider offering GPU instances and clusters for training and inference, plus on-premises hardware, aimed at AI researchers and teams needing GPUs.

$$ Midusage-based pricing
Strengths
Maturity, operating 14 yearsPrice, Moderate hourly cost

Modal

modal.com

Modal is a serverless platform for running compute-intensive and GPU workloads directly from Python code without managing infrastructure, aimed at developers running AI and batch jobs.

$$ Midusage-based pricing
Strengths
Free tier, Free plan availableMaturity, operating 5 yearsPrice, Moderate hourly cost

Nebius

nebius.com

Nebius is an AI-focused cloud provider offering GPU infrastructure and managed services for AI workloads, aimed at AI teams needing scalable compute.

$$ Midusage-based pricing
Strengths
Price, Moderate hourly costValue, Mid tier

Paperspace

www.paperspace.com

Paperspace, now part of DigitalOcean, provides GPU cloud instances and notebooks for machine learning and development, aimed at developers and ML practitioners.

$$$ Premiumusage-based pricing
Strengths
Maturity, operating 12 yearsPrice, Premium hourly cost

RunPod

www.runpod.io

RunPod is a GPU cloud offering both on-demand and serverless GPU compute for AI workloads, aimed at developers wanting affordable, flexible GPU access.

$$ Midusage-based pricing
Strengths
Flexibility, Spot instances +2Price, Moderate hourly cost

TensorDock

tensordock.com

TensorDock is a GPU cloud marketplace aggregating capacity from many host providers, offering a wide range of GPU types at market-driven prices.

$ Budgetusage-based pricing
Strengths
Maturity, operating 5 yearsValue, Budget tier