Providers
Every cloud we price, and how they differ
Hyperscalers charge for breadth and an SLA. GPU-native clouds charge for the GPU. Marketplaces charge whatever a stranger with a rack will accept. The scores below are how the comparison engine weighs that difference.
AWS
HyperscalerDeepest region coverage and the widest managed-service catalog. You pay a premium for both.
- Reliability
- 96/100
- Provisioning
- 72/100
- Egress
- $0.090/GB
48 offers · from $1.006/hr
DetailsGoogle Cloud
HyperscalerStrong networking and TPU alternatives. Committed-use discounts reward predictable workloads.
- Reliability
- 94/100
- Provisioning
- 74/100
- Egress
- $0.085/GB
48 offers · from $0.916/hr
DetailsMicrosoft Azure
HyperscalerEnterprise agreements and hybrid identity. Typically the priciest per GPU-hour of the big three.
- Reliability
- 93/100
- Provisioning
- 68/100
- Egress
- $0.087/GB
48 offers · from $1.149/hr
DetailsVast.ai
MarketplaceOpen marketplace of third-party hosts. Cheapest per GPU-hour, most variable in reliability.
- Reliability
- 58/100
- Provisioning
- 86/100
- Egress
- Free
48 offers · from $0.278/hr
DetailsCoreWeave
GPU cloudGPU-native cloud with InfiniBand fabrics. Fast multi-node scheduling at well under hyperscaler rates.
- Reliability
- 84/100
- Provisioning
- 88/100
- Egress
- $0.010/GB
42 offers · from $0.734/hr
DetailsLambda Labs
GPU cloudSimple flat pricing and preinstalled ML images. Capacity for the newest GPUs sells out quickly.
- Reliability
- 82/100
- Provisioning
- 80/100
- Egress
- Free
28 offers · from $0.605/hr
DetailsRunPod
GPU cloudPer-second billing and instant pod start. Community cloud tiers undercut everyone but AWS spot.
- Reliability
- 74/100
- Provisioning
- 92/100
- Egress
- Free
48 offers · from $0.400/hr
DetailsPrices here are entry points. Run your actual workload through the calculator to see what it costs on each.
Price a workload