GPU Pricing 2026: Verified H100 Cost Per Hour from $1.33 to $12.29 Across 17 Cloud Providers, No Affiliate Links

The business is an independent GPU cloud pricing reference that publishes vendor-neutral hourly rates for H100, A100, and L40S GPUs across 17 providers. It lists specialist clouds like CoreWeave and Lambda Labs alongside hyperscalers such as AWS, Azure, and GCP. The site draws only from publicly listed vendor prices and was last verified in July 2026.

Small teams with 1 to 4 GPUs typically spend between $500 and $5,000 per month on inference and fine-tuning workloads. At this scale, pay-by-the-second models from Vast.ai marketplace, RunPod Community Cloud, or Modal serverless are often cheaper than reserved options. The site notes that marketplace models clear capacity from independent hosts but may involve interruptibility.

Mid-sized clusters requiring 8 to 64 H100 or A100 GPUs range from $15,000 to $250,000 per month. Providers like CoreWeave, Lambda Reserved, Crusoe, and Together AI offer reserved-capacity contracts for these sizes. These contracts typically run for 1 to 3 years and unlock discounts of 30 to 60 percent off list rates.

Large-scale deployments involving 256 or more H100, H200, or B200 GPUs cost between $1M and $40M+ per month. These setups often use hyperscaler SuperPODs, CoreWeave HGX clusters, or Oracle bare-metal SuperClusters. They rely on multi-year reservations and InfiniBand-first networking to manage the workload.

The site provides a GPU Cloud Cost Calculator where users can pick a GPU class, set cluster size, and define duty cycle. Users can also add a 25 percent overhead factor for storage, egress, and MLOps. The tool shows the range from the cheapest published rate to the highest hyperscaler list rate.

Hyperscaler list rates are positioned for spot-buyers and represent the highest tier offered by AWS, Azure, and GCP. Actual prices for enterprise customers under Savings Plans or Reserved Instances are materially lower. The integration premium includes IAM, S3, EFA, SageMaker, and Foundry, which specialist clouds do not match.

Per-second billing is generally cheaper than per-hour billing for inference workloads that process only a few requests per minute. However, continuous training workloads often price at a premium under per-second models. Cold-start billing on serverless platforms is a specific gotcha to consider.

Specialist clouds like CoreWeave, Lambda, Together, Crusoe, and DigitalOcean typically publish 30 to 60 percent off on-demand rates for 1 to 3 year reservations. Hyperscalers offer similar 30 to 50 percent discounts via Reserved Instances or Savings Plans. Enterprise EA contracts at all three hyperscalers add a further committed-spend discount.

The site covers Nvidia GPUs including the H100, H200, B200, A100, L40S, A10, A10G, and T4. It specifically excludes TPUs and other accelerators from its current scope. The methodology involves pulling rates from public pricing pages and normalizing them per GPU per hour.

The site carries no affiliate links or sponsored placements to maintain independence from the vendors listed. If a commercial link is added in the future, it will be clearly labelled. The goal is to provide a single page for AI infrastructure buyers to scope a realistic budget before speaking to a vendor.

Top-traffic vendors such as CoreWeave, Lambda, RunPod, AWS, Azure, and GCP are re-verified monthly. The long tail of providers like Crusoe, Hyperstack, and Together AI is re-verified quarterly. Every page footer carries the most recent verified date to track changes.

The site was founded by Oliver Wakefield-Smith, who is the Founder of Digital Signet. Digital Signet operates a portfolio of vendor-neutral cost references. The site is part of a larger network of adjacent AI infrastructure categories including embeddings and vector databases.

The site offers a 2026 GPU pricing guide, a pricing calculator, and grouped recommendations for small teams, mid-sized clusters, and large-scale deployments. It also provides an RFP template and implementation cost guides. The methodology page details how rates are sampled and re-verified on a recurring schedule.

The official website of GPU Pricing 2026: Verified H100 Cost Per Hour from $1.33 to $12.29 Across 17 Cloud Providers, No Affiliate Links is gpucloudcost.com.

GPU Pricing 2026: Verified H100 Cost Per Hour from $1.33 to $12.29 Across 17 Cloud Providers, No Affiliate Links has no reviews on Trustburn yet. Anyone who has used it can write the first review on its Trustburn page.