GPU Cloud Provider Comparison Matrix

Compare 15+ GPU Cloud Providers for AI Training & Inference

1What It Measures

Side-by-side comparison of GPU cloud providers including hyperscalers (AWS, Azure, GCP), neoclouds (CoreWeave, Lambda), and marketplaces (Vast.ai, RunPod). Pricing updated weekly from live sources.

2Why It Matters

Choosing the right GPU cloud provider can save 50-70% on AI infrastructure costs. Different providers optimize for different workloads - training vs inference, research vs production, compliance vs cost.

3How to Read It

Filter by provider tier, sort by price or name. Click any provider for detailed comparison pages. Use the GPU Rental Cost Calculator for precise cost estimates.

Key Metrics Explained
Providers Tracked
15

H100 cloud options

Cheapest H100
$1.89/hr

Vast.ai

Price Range
3.9x

Cheapest to most expensive

Calculate Your Costs →
PRO

Get Provider Availability Alerts & Spot Pricing

PRO members get real-time alerts when providers add capacity, spot pricing intelligence, and negotiated enterprise rates.

  • ✓ Real-time H100/H200 availability tracking
  • ✓ Spot pricing alerts (save 50-70%)
  • ✓ SLA comparison & contract analysis tools
  • ✓ Custom provider export (PDF/CSV)
Upgrade to PRO

Vast.ai

marketplace
$1.89
per H100/hr
$1.89
H100 SXM
N/A
H200
$0.89
A100

Batch processing

DePIN60-90% cheaperBatch only
Pros
  • • Lowest prices
  • • Large selection
Cons
  • • No guarantees
  • • Variable quality
Regions: Global P2P

CoreWeave

neocloud
$1.99
per H100/hr
$1.99
H100 SXM
$2.49
H200
$1.21
A100

AI training at scale

KubernetesInfiniBandSpot pricing
Pros
  • • Best price/performance
  • • Production-grade
Cons
  • • Kubernetes required
  • • No SageMaker equivalent
Regions: us-east, us-west, uk

Voltage Park

neocloud
$2.1
per H100/hr
$2.1
H100 SXM
$2.6
H200
$1.2
A100

Large training runs

Large clustersDedicatedEnterprise
Pros
  • • Large clusters
  • • Dedicated capacity
Cons
  • • Weekly commitment
  • • US only
Regions: us

Nebius

neocloud
$2.2
per H100/hr
$2.2
H100 SXM
$2.7
H200
$1.25
A100

International research

OrchestrationInfiniBandResearch
Pros
  • • Orchestration tools
  • • InfiniBand
Cons
  • • EU/Israel only
  • • Smaller ecosystem
Regions: eu, israel

Crusoe Energy

neocloud
$2.3
per H100/hr
$2.3
H100 SXM
$2.8
H200
$1.35
A100

Green AI workloads

Stranded gasCarbon neutralDedicated
Pros
  • • Carbon neutral
  • • Stranded gas power
Cons
  • • Monthly commitment
  • • Limited regions
Regions: us-wy, us-tx

RunPod

marketplace
$2.39
per H100/hr
$2.39
H100 SXM
N/A
H200
$1.09
A100

Serverless inference

ServerlessSpot GPUsTemplates
Pros
  • • Serverless option
  • • Global coverage
Cons
  • • Variable quality
  • • No SLA
Regions: 31+ regions

Lambda Labs

neocloud
$2.49
per H100/hr
$2.49
H100 SXM
$2.99
H200
$1.29
A100

ML research teams

1-Click ClustersPyTorch nativeResearch focus
Pros
  • • Easiest setup
  • • Research-friendly
Cons
  • • Limited regions
  • • No spot pricing
Regions: us-tx, us-az, us-ut

Genesis Cloud

neocloud
$2.69
per H100/hr
$2.69
H100 SXM
N/A
H200
$1.59
A100

Green/ESG workloads

100% renewableEU GDPRESG
Pros
  • • 100% renewable
  • • EU GDPR
Cons
  • • Iceland only
  • • Latency for US
Regions: eu-iceland

Hyperstack

neocloud
$2.85
per H100/hr
$2.85
H100 SXM
N/A
H200
$1.65
A100

EU enterprise

Enterprise SLAGDPRConsulting
Pros
  • • Enterprise SLA
  • • GDPR compliant
Cons
  • • Higher pricing
  • • Limited GPU types
Regions: uk, canada

Google Cloud

hyperscaler
$3
per H100/hr
$3
H100 SXM
N/A
H200
$3.67
A100

Google AI stack users

Vertex AITPU accessEnterprise
Pros
  • • Gemini integration
  • • TPU option
Cons
  • • No InfiniBand
  • • Limited H100 availability
Regions: us-central1, us-east4, europe-west4

Paperspace

marketplace
$3.09
per H100/hr
$3.09
H100 SXM
N/A
H200
$1.89
A100

ML development

Gradient MLNotebooksDev-friendly
Pros
  • • Gradient ML
  • • Notebooks
Cons
  • • Higher pricing
  • • Limited H100s
Regions: us-east, us-west, eu

Oracle Cloud

hyperscaler
$3.5
per H100/hr
$3.5
H100 SXM
$4.25
H200
$2.95
A100

OpenAI workloads

Bare metalRDMAOpenAI partner
Pros
  • • Bare metal
  • • Good pricing
Cons
  • • Annual commitment
  • • Smaller ecosystem
Regions: us-ashburn-1, us-phoenix-1, eu-frankfurt-1

AWS

hyperscaler
$3.9
per H100/hr
$3.9
H100 SXM
N/A
H200
$3.97
A100

Enterprise production

SageMakerEnterprise SLACompliance
Pros
  • • Enterprise compliance
  • • Global reach
Cons
  • • 2x more expensive
  • • Complex pricing
Regions: us-east-1, us-west-2, eu-west-1

IBM Cloud

hyperscaler
$5.2
per H100/hr
$5.2
H100 SXM
N/A
H200
$3.5
A100

IBM ecosystem users

ComplianceHybrid cloudwatsonx
Pros
  • • Enterprise compliance
  • • watsonx integration
Cons
  • • Higher pricing
  • • Smaller GPU fleet
Regions: us-south, eu-de, jp-tok

Microsoft Azure

hyperscaler
$7.35
per H100/hr
$7.35
H100 SXM
N/A
H200
$3.4
A100

Microsoft shops

Azure MLEnterpriseCompliance
Pros
  • • O365 integration
  • • Azure credits
Cons
  • • Most expensive
  • • Complex
Regions: eastus, westus2, westeurope

Need Help Choosing?

Use our GPU Rental Cost Calculator to compare exact costs for your workload.

Calculate GPU Costs →