Show HN: The cheapest GPU cloud – H100s at $2.04/HR, H200s at $3/HR
Summary
The article presents Compute Cheap's pricing for NVIDIA H100 and H200 GPUs, detailing on-demand and spot prices and how frontier AI workloads can be run at significantly lower costs. It explains the economics behind the pricing—competition, idle capacity, prepaid blocks, and a simplified stack—along with a call to action to request compute capacity. Daily price comparisons and charts illustrate the cost advantages for training and inference workloads.