Amazon Web Services (AWS)
AvailableBest for Enterprise Production, Model Deployment, Massive Scale
Locations: Global (30+ regions)
$1.00/hr
View ProviderGPU Type
NVIDIA H100 GPU Cloud Providers
Compare 20 cloud providers offering NVIDIA H100 (80 GB HBM3 VRAM). Find real-time pricing, availability, and get matched with verified providers instantly.
Memory80 GB HBM3
FP16 Performance989 TFLOPS
Memory Bandwidth3.35 TB/s
Providers Listed20+
What is the NVIDIA H100?
The NVIDIA H100 is the most powerful data-center GPU available, purpose-built for accelerating transformer-based large language models. With 80 GB of HBM3 memory and NVLink 4.0 support, the H100 is the gold standard for training GPT-class models at scale.
Primary Use Cases for NVIDIA H100
- LLM training
- large-scale AI research
- multi-modal model training
Pricing Benchmark for NVIDIA H100 in 2026
The spot market for H100 cloud compute varies widely by provider. On-demand pricing typically ranges from $1.50–$5/hr per GPU for single-instance access. For larger multi-GPU clusters (8x, 16x, or 64x GPU nodes), enterprise pricing with SLAs is negotiated directly with providers. Reserved capacity offers 30–60% discounts vs. on-demand pricing.
How to Choose a NVIDIA H100 Cloud Provider
When evaluating providers for H100 GPU cloud, consider:
- Pricing model: On-demand vs. reserved vs. spot pricing
- Cluster size: Maximum number of GPUs per node and interconnect bandwidth
- Storage: NVMe SSD throughput and object storage availability
- Networking: InfiniBand vs. Ethernet interconnect for distributed training
- Location: Data residency requirements and latency to your users
- Support SLA: Response time guarantees and dedicated support options
20 Cloud Providers Offering NVIDIA H100
CoreWeave
AvailableBest for Enterprise LLM Training, HPC, AI Inference at Scale
Locations: US East (NJ, VA), US West (CA), EU West (UK, Sweden, Netherlands)
$0.99/hr
View Provider
Google Cloud (GCP)
AvailableBest for AI Innovation, TPU Training, MLOps (Vertex AI)
Locations: Global (35+ regions)
$1.00/hr
View Provider
Together AI
AvailableBest for Finetuning Open Source Models, Serverless inference endpoints
Locations: US, EU
$0.20/hr
View Provider
DeepInfra
AvailableBest for LLM Serverless APIs, Fast Image Generation, Voice AI
Locations: US East, US West
$0.15/hr
View Provider
Lambda Labs
AvailableBest for LLM Training, AI Research, Fine-Tuning
Locations: US East (Texas), US West (California, Utah), Europe (UK)
$0.50/hr
View Provider
Microsoft Azure
AvailableBest for Enterprises, OpenAI Integrations, Hybrid Cloud
Locations: Global (60+ regions)
$1.00/hr
View Provider
OctoAI
AvailableBest for Production AI Model Serving, Custom Model Inference
Locations: US East, US West
$0.20/hr
View Provider
Replicate
AvailableBest for Serverless Image Generation, LLM API inference, Open-Source Model Hosting
Locations: US, EU
$0.36/hr
View Provider
Anyscale
AvailableBest for Distributed Computing, Ray workload scaling, LLM hosting
Locations: US East, US West
$0.50/hr
View Provider
Modal
AvailableBest for Serverless Inference, Ad-hoc Python scripts, Quick Prototyping
Locations: US East, US West, Europe
$0.50/hr
View Provider
$1.50/hr
View Provider
Baseten
AvailableBest for Scale-to-zero Inference, Custom Model Serving, Low-Latency APIs
Locations: US, EU
$0.40/hr
View Provider
RunPod
AvailableBest for AI Inference, Image Generation, Fine-Tuning, Budget ML
Locations: US East, US West, EU West (Norway, France), Asia Pacific (Singapore)
$0.14/hr
View Provider
Vultr
AvailableBest for Edge AI, Application Developers requiring unified infrastructure, Web Apps + AI
Locations: Global (30+ Data Centers)
$0.20/hr
View Provider
$0.80/hr
View Provider
Nebius AI
AvailableBest for European Enterprise AI, Massive Scale LLM Training, HPC
Locations: EU (Finland)
$2.50/hr
View Provider
Voltage Park
LimitedBest for Enterprise LLM Pre-training, Large-Scale AI Research, Foundation Model Development
Locations: US West (Colorado, Nevada), US East (Virginia)
$2.25/hr
View Provider
FluidStack
AvailableBest for Enterprise AI Training, Multi-Tenant GPU Clusters, Cost-Effective H100 Access
Locations: UK (London, Manchester), US West (California), US East (Virginia), EU Central (Germany, France)
$0.89/hr
View Provider
Paperspace by DigitalOcean
AvailableBest for ML Notebooks, AI Model Development, Research, Computer Vision
Locations: US East (New York), US West (California), EU West (Netherlands, UK)
$0.08/hr
View ProviderFrequently Asked Questions
NVIDIA H100 is commonly used for: LLM training, large-scale AI research, multi-modal model training. Its 80 GB HBM3 of VRAM makes it suitable for running large models that don't fit in smaller GPU memory.
NVIDIA H100 cloud pricing varies by provider and region, but typically ranges from $1.50/hr to $8/hr for single-GPU instances. Multi-GPU cluster pricing scales proportionally. Use the filters above to compare current market rates.
ComputeStacker currently lists 20 providers offering H100 GPU cloud access. These include a mix of hyperscalers, specialist AI cloud providers, and bare-metal GPU hosting services.
Yes — most providers on ComputeStacker offer on-demand hourly pricing for H100 instances. Reserved and spot pricing options are also available from many providers, offering discounts of 30–70% for committed usage.