News
DealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanSovereignNeysa Deploys AMD GPU Cluster in Mumbai for AI Training and InferenceInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to NvidiaDealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanSovereignNeysa Deploys AMD GPU Cluster in Mumbai for AI Training and InferenceInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to Nvidia
EveryNeocloud

GPU virtual machines

GPU instances by the hour. 96 providers on EveryNeocloud so far.

A GPU virtual machine (VM) is a slice of a server, with one or more GPUs attached, that a customer can start in minutes and pay for by the hour or second. It is the standard way clouds sell compute. AWS EC2, Google Compute Engine and Azure virtual machines offer GPU types from small inference cards to eight-GPU H100 and Blackwell machines.

How it works

A hypervisor divides the server and passes GPUs straight through to each virtual machine, so performance is close to bare metal. Customers choose an instance type, a fixed mix of GPUs, CPUs, memory and storage, and can stop it when done. Some GPUs can be split further: Nvidia's Multi-Instance GPU cuts one chip into up to seven parts, so small jobs pay less.

Prices and limits

On-demand prices are the highest; reserved instances, committed for one to three years, cost less; spot instances, spare capacity that can be taken back, cost least. In 2023 and 2024 the largest GPU instances were often sold out, and clouds asked customers to reserve ahead. As supply grew in 2025, hourly prices for H100 machines fell across most providers.

Famous GPU virtual machines providers

Further reading

Photo credits on this page

The small pictures above come from Wikimedia Commons under free licences. Some are cropped to fit.

  • Amazon Elastic Compute Cloud: Raysonho @ Open Grid Scheduler / Grid Engine, CC BY-SA 3.0, source

The weekly digest

Every Monday: the provider of the day, new pages on EveryNeocloud, and the conferences and anniversaries of the week ahead.