AI compute market signals and learning

Providers

AI GPU Provider Directory

A curated directory of public AI GPU price pages and comparable list-price signals.

Use this as a curated public-price directory, not a quote engine. Every row links to a provider price page and shows list price, H100e-normalized price where applicable, availability, and how long the current list price has held.

Providers

All Providers Directory

8 editorially checked provider rows. Use the provider-type filter to narrow the directory.

AWSHyperscalersEnterprise cloud baselineH100us-east-1$6.88$12.29Quota Gated23 daysOpen price page
AzureHyperscalersMicrosoft enterprise baselineH100eastus$12.29$12.29Quota Gated59 daysOpen price page
CrusoeAI-native cloudsEnergy-sited managed clusterH100Multi-region$3.90$3.90Available Limited47 daysOpen price page
CrusoeAI-native cloudsEnergy-sited managed clusterH200Multi-region$4.29$3.73Available32 daysOpen price page
LambdaDeveloper GPU cloudsDeveloper self-serve GPUB200Multi-region$6.99$3.50Available32 daysOpen price page
LambdaDeveloper GPU cloudsDeveloper self-serve GPUH100Multi-region$4.29$4.29Available47 daysOpen price page
RunPod SecureDeveloper GPU cloudsBurst or dev Secure CloudB200Multi-region$5.89$2.94Available32 daysOpen price page
RunPod SecureDeveloper GPU cloudsBurst or dev Secure CloudH200Multi-region$4.39$3.82Available32 daysOpen price page

Rows are public on-demand list prices unless noted. Provider links open ComputeTape profiles; public price page links open the original source.

Buyer checklist

How to compare these rows

Line up comparable rows before treating one provider as cheaper or better suited.

  • Match GPU, access mode, and region before comparing hourly rates.
  • Separate public list price from negotiated, reserved, capacity-block, or spot terms.
  • Use the source link, provider profile, and No change in streak to judge evidence quality.
Learn more about comparing

About the list

Source and freshness standards

Rows appear only after editorial review and stay within the recheck window. Unreviewed or stale rows stay out of the public directory.

Evidence labels reference

Rights, freshness, and confidence labels for every row.

Use the directory

Turn provider information into a plan

GPU Pricing Hub

The per-chip price bands and the current H100 read built from these same reviewed rows.

GPU-Hour Cost Calculator

Estimate AI compute cost from GPU price, runtime, utilization, and overhead.

Reserved vs On-Demand Calculator

Compare a GPU reservation to pure on-demand at expected usage.