What is the cheapest NVIDIA A100 GPU cloud provider?

Thunder Compute offers the lowest publicly listed NVIDIA A100 80GB price in our June 2026 comparison at $1.09 per GPU-hour. Other low-cost providers include Runpod at $1.39/hr and Vast.ai at $1.94/hr.

How much does an NVIDIA A100 GPU cost on AWS, Azure, and Google Cloud?

AWS charges approximately $3.43 per GPU-hour for A100 GPUs on p4de.24xlarge instances after normalizing the 8-GPU node price. Azure lists NC24ads A100 v4 instances at $4.41/hr, while Google Cloud's a2-ultragpu-1g costs $5.07/hr.

How much more does an NVIDIA A100 80GB cost compared with the 40GB model?

Pricing differences between A100 40GB and A100 80GB GPUs vary by provider. In June 2026, publicly listed A100 80GB prices range from $1.09/hr to $5.07/hr. Because many providers no longer publish A100 40GB pricing separately, the premium depends on availability and region.

Are NVIDIA A100 GPU prices the same across all regions?

No. A100 pricing varies by region even within the same cloud provider. The figures in this comparison reflect U.S. pricing, while rates in other regions may differ by 5–20%.

How much can I save using budget GPU providers instead of AWS, Azure, or Google Cloud?

Lower-cost GPU clouds can significantly reduce compute expenses. A 10-hour workload costs $10.90 on Thunder Compute, compared with about $34.30 on AWS, $44.10 on Azure, and $50.70 on Google Cloud.

What should I know about marketplace GPU providers like Vast.ai and Runpod?

Marketplace-based platforms such as Vast.ai and Runpod often provide lower pricing than hyperscalers. In June 2026, Vast.ai averaged about $1.94 per GPU-hour and Runpod listed A100 instances at $1.39/hr. Pricing and reliability can vary because compute capacity is supplied by multiple hosts.

Which major cloud provider has the lowest A100 pricing?

Among the major hyperscalers, AWS offers the lowest publicly listed A100 pricing at approximately $3.43 per GPU-hour. Azure follows at $4.41/hr, while Google Cloud charges about $5.07/hr.

What is the hourly price of AWS p4de.24xlarge A100 instances?

The AWS p4de.24xlarge instance costs $27.44 per hour and contains eight NVIDIA A100 GPUs, resulting in an effective cost of approximately $3.43 per GPU-hour.

Where can I rent NVIDIA A100 GPUs for short-term projects?

Providers including Thunder Compute, Runpod, Vast.ai, Crusoe Cloud, Lambda, Paperspace, AWS, Azure, and Google Cloud offer on-demand A100 instances with hourly billing, making them suitable for development, fine-tuning, and temporary AI workloads.

What is the typical price range for NVIDIA A100 cloud instances?

Public on-demand A100 pricing ranges from $1.09 to $5.07 per GPU-hour in June 2026. Specialized GPU clouds are generally much cheaper than major hyperscalers while still providing immediate access without hardware procurement delays.

Go back

NVIDIA A100 Pricing (July 2026): Cheapest On-demand GPU Instances

Carl PetersonJuly 1, 20263 min read

The A100 GPU rental price starts at $1.09 per GPU-hour on Thunder Compute in this June 2026 comparison of on-demand NVIDIA A100 cloud GPU pricing.

This article compares cloud rental rates, not the hardware purchase price of an NVIDIA A100 GPU. Below is a price snapshot for A100 GPU rental by provider. All numbers are on-demand, pay-by-the-hour in a U.S. region unless noted.

Provider	Instance	On-Demand $/GPU-hr*	Notes
Thunder Compute		$1.09	On-demand
Runpod		$1.39	Community marketplace rate; varies by supply
Crusoe Cloud		$2.00	Public listed hourly pricing
Vultr		$2.40	Public listed hourly pricing
CoreWeave	8-GPU node	$2.50	Normalized from public 8-GPU node pricing
Modal		$2.50	Public listed hourly pricing
Vast.ai		$1.94	Average marketplace pricing.
Lambda		$2.79	Flat per-GPU rate
Paperspace		$3.18	Public listed price
AWS EC2	p4de.24xlarge	$3.43	$27.44/hr VM ÷ 8 GPUs
Azure	NC24ads A100 v4	$4.41	Public listed single-GPU VM
Oracle Cloud	BM.GPU.A100-v2.8	$4.00	8-GPU bare-metal node, normalized per GPU
Google Cloud	a2-ultragpu-1g	$5.07
Nebius		Not listed	Public pricing page doesn't list A100 instances

Single-GPU costs are shown for multi-GPU nodes for easy comparison.
Network, storage, and egress fees are not included.
Last update: July 1, 2026.

Thunder Compute is about 6.5x cheaper than Google Cloud for a single A100, about 4.4x cheaper than AWS, and about 2.1x cheaper than Crusoe Cloud. Across the listed providers, alternatives range from about 1.2x to 6.5x Thunder Compute's hourly price.

Cost to Run an A100 for 2 Hours

Provider	On-Demand $/GPU-hr	2-Hour Cost
Thunder Compute	$1.09	$2.18
Runpod	$1.39	$2.78
Crusoe Cloud	$2.00	$4.00
Vast.ai	$1.94	$3.88
Vultr	$2.40	$4.80
CoreWeave	$2.50	$5.00
Modal	$2.50	$5.00
Lambda	$2.79	$5.58
Paperspace	$3.18	$6.36
AWS EC2	$3.43	$6.86
Oracle Cloud	$4.00	$8.00
Azure	$4.41	$8.82
Google Cloud	$5.07	$10.14
Nebius	Not listed	Not listed

Methodology

On-demand only: No reserved-instance, commitment, or prepaid discounts.
Same class of silicon: Listed rates are for NVIDIA A100 80 GB GPUs where the provider publishes A100 pricing.
Public price lists: Each figure comes from the vendor's current pricing page on the date above.
USD in U.S. regions: Rates in other regions can differ by 5-20%.
We update this data every month.

A100 Cost for Longer Workloads

Even small differences in hourly pricing add up quickly. For example, a 100-hour A100 workload costs $109.00 on Thunder Compute, $200.00 on Crusoe Cloud, $343.00 on AWS, and $507.00 on Google Cloud, before storage, networking, or egress fees.

See how different platforms stack up against each other in our breakdown of the cheapest cloud GPU providers.