Cloud GPU Providers

Compare pricing and availability across cloud platforms

17
Cloud Providers
4
Hyperscalers
9
GPU-Native Clouds
4
Marketplaces

Provider Price Comparison

Compare GPU cloud pricing across all providers. Prices shown are per-GPU hourly rates.

Provider Type GPUs Lowest Price H100 Price A100 Price Spot Available
TensorDock Marketplace 9 $0.060/hr $2.25/hr - No View →
Vast.ai Marketplace 4 $0.081/hr - - Yes (-23%) View →
RunPod GPU Cloud 41 $0.130/hr $2.39/hr $1.39/hr Yes (-54%) View →
Datacrunch (Verda) GPU Cloud 9 $0.140/hr $2.29/hr - No View →
Genesis Cloud GPU Cloud 4 $0.250/hr $2.19/hr - No View →
Google Cloud Platform Enterprise 10 $0.350/hr $9.80/hr - Yes (-60%) View →
CoreWeave GPU Cloud 12 $0.390/hr $4.25/hr - No View →
Amazon Web Services Enterprise 8 $0.420/hr - $4.10/hr Yes (-57%) View →
Microsoft Azure Enterprise 6 $0.454/hr $12.29/hr $3.40/hr No View →
Jarvis Labs GPU Cloud 6 $0.490/hr $2.99/hr - No View →
Lambda Labs GPU Cloud 11 $0.500/hr $2.49/hr - No View →
Paperspace GPU Cloud 7 $0.760/hr $2.24/hr - No View →
Oracle Cloud Enterprise 3 $1.00/hr $10.75/hr - No View →
Fluidstack GPU Cloud 5 $1.30/hr $2.10/hr - No View →

Which Provider Should You Choose?

The best cloud GPU provider depends on your specific needs. Here are our recommendations:

Budget-Conscious

Best for hobbyists, students, and cost-sensitive projects.

TensorDock from $0.060/hr Vast.ai from $0.081/hr
Enterprise & Compliance

SLAs, compliance certifications, and integration with existing cloud infrastructure.

Oracle Cloud Google Cloud Platform Microsoft Azure
AI/ML Training

Optimized for large-scale training with H100/A100 availability.

Fluidstack H100 $2.10/hr Genesis Cloud H100 $2.19/hr
Quick Prototyping

Fast setup, pay-as-you-go, no long-term commitment.

RunPod Paperspace
Pro Tip: The cheapest H100 is currently available at Vast.ai for $1.38/hr.

Hyperscalers 4 providers

Oracle Cloud
Enterprise
3 GPUs H100: $10.75/hr
from $1.00/hr
Google Cloud Platform
Enterprise
10 GPUs H100: $9.80/hr
from $0.350/hr Spot Available
Microsoft Azure
Enterprise
6 GPUs H100: $12.29/hr A100: $3.40/hr
from $0.454/hr
Amazon Web Services
Enterprise
8 GPUs A100: $4.10/hr
from $0.420/hr Spot Available

GPU-Native Clouds 9 providers

Genesis Cloud

欧洲 GPU 云服务商

0 GPUs
RunPod
Spot
41 GPUs H100: $2.39/hr A100: $1.39/hr
from $0.130/hr Spot Available
Paperspace
Cheap H100
7 GPUs H100: $2.24/hr
from $0.760/hr
Lambda Labs
Cheap H100
11 GPUs H100: $2.49/hr
from $0.500/hr
Jarvis Labs
Cheap H100
6 GPUs H100: $2.99/hr
from $0.490/hr
Genesis Cloud
Cheap H100
4 GPUs H100: $2.19/hr
from $0.250/hr
Fluidstack
Cheap H100
5 GPUs H100: $2.10/hr
from $1.30/hr
Datacrunch (Verda)
Cheap H100
9 GPUs H100: $2.29/hr
from $0.140/hr
CoreWeave
Wide Selection
12 GPUs H100: $4.25/hr
from $0.390/hr

GPU Marketplaces 4 providers

Vast.ai

P2P GPU 市场，价格最低

0 GPUs
Massed Compute

专注 AI 训练的 GPU 云

0 GPUs
Vast.ai
Spot
4 GPUs
from $0.081/hr Spot Available
TensorDock
Cheap H100
9 GPUs H100: $2.25/hr
from $0.060/hr

How to Choose a Cloud GPU Provider

Hyperscalers

AWS, Azure, GCP - Best for enterprise workloads requiring SLAs, compliance (SOC2, HIPAA), and integration with existing cloud infrastructure. Higher prices but guaranteed availability and support.

  • ✓ Enterprise SLAs
  • ✓ Compliance certifications
  • ✓ Global availability
  • ✗ Premium pricing
GPU-Native Clouds

CoreWeave, Lambda, Paperspace - Specialized for AI/ML with competitive pricing and better GPU availability. Great balance of price, reliability, and features.

  • ✓ AI-optimized infrastructure
  • ✓ Better H100/A100 availability
  • ✓ Competitive pricing
  • ✓ ML-focused tooling
Marketplaces

Vast.ai, RunPod - Lowest prices through P2P or aggregated supply. Best for cost-sensitive workloads that can tolerate occasional interruptions.

  • ✓ Lowest prices (up to 80% cheaper)
  • ✓ Flexible spot pricing
  • ✓ Wide GPU selection
  • ✗ Variable availability
Key Factors to Consider
Budget: Marketplaces offer 50-80% savings
Reliability: Hyperscalers for mission-critical
GPU Type: Check H100/A100 availability
Scale: Reserved pricing for long-term

Cloud GPU Market Insights

Hyperscalers
$3.32/hr
Average GPU Price
GPU Clouds
$1.48/hr
Average GPU Price
Marketplaces
$0.666/hr
Average GPU Price
Marketplaces are on average 80% cheaper than hyperscalers

Frequently Asked Questions

Marketplace providers like Vast.ai and RunPod typically offer the lowest prices, often 50-80% cheaper than hyperscalers. However, availability can be variable. For reliable low-cost options, Lambda Labs and Paperspace offer competitive pricing with better availability guarantees.

CoreWeave is generally better for dedicated AI/ML workloads with better H100/A100 availability and lower prices. AWS is better if you need integration with other AWS services, enterprise compliance, or global availability. CoreWeave can be 30-50% cheaper for equivalent GPU instances.

Spot instances can save 60-90% but come with interruption risk. They're ideal for: fault-tolerant training with checkpointing, hyperparameter tuning, inference testing, and batch processing. Avoid for: production inference, time-sensitive training, or workloads without checkpoint support.

GPU-native clouds like CoreWeave, Lambda Labs, and Crusoe typically have better H100 availability than hyperscalers. Hyperscalers often have long waitlists for H100 instances. For immediate access, check marketplace providers, though availability varies.

Use on-demand for: short-term projects, variable workloads, testing. Use reserved (1-3 year) for: production workloads, predictable usage, cost optimization (save 30-60%). Most providers offer reserved pricing for commitments of 1 month or longer.

Hyperscalers (AWS, Azure, GCP) offer broad cloud services with GPUs as one option - better for enterprises needing compliance and integration. GPU-native clouds (CoreWeave, Lambda) specialize in GPU compute - better pricing, availability, and ML-focused features for AI workloads.