How AM ThermaLink compares

A simple side-by-side of the cheapest public GPU option at each company — the most VRAM value per hour, with AM ThermaLink kept in the first column for easy scanning.

Cheapest per VRAM option

The lowest public GPU option that gives the most VRAM value per hour for each company.

$/GB-hr
AM ThermaLink AWS Google Azure CoreWeave Lambda RunPod
GPU Intel Arc Pro B70Budget high-VRAM tier NVIDIA L4 / A10GClosest public low-cost rows NVIDIA L4G2-class GPU NVIDIA T4Closest low-cost public row NVIDIA L4048GB public row NVIDIA A600048GB public row RTX A5000Lowest public $/GB row
VRAM 32GB 24GB 24GB 16GB 48GB 48GB 24GB
Price $0.26-$0.339/ GPU-hr ~$0.80-$1.01/ GPU-hr ~$0.70/ GPU-hr ~$0.526/ VM-hr ~$1.25/ GPU-hr $1.09/ GPU-hr $0.27/ GPU-hr
Approx $/GB $0.008-$0.011Cheapest $/GB ~$0.033-$0.042 ~$0.029 ~$0.033 ~$0.026 ~$0.023 ~$0.011

Prices are approximate public on-demand/list figures reviewed in July 2026 and normalized where possible. They are not quotes.

Cloud totals often add region, reservation, operating system, networking, egress, request, and performance-tier charges. AM ThermaLink pricing remains marked as early placeholder pricing elsewhere on this site until finalized.

Get a number for your exact workload

Use the estimator to combine compute and storage for your own configuration, or pre-order to lock in early-access pricing.