AWS Hikes EC2 Capacity Block Rates by 15% in Uniform ML Pricing Adjustment

"AWS has increased pricing for EC2 Capacity Blocks for ML by approximately 15% across all regions where the service is available. The price adjustment affects organizations reserving dedicated GPU capacity for large-scale machine learning workloads, with rates rising uniformly across AWS's most powerful ML instances, including P5en, P5e, P5, and P4d, powered by NVIDIA GPUs, as well as Trn2 and Trn1 instances that use AWS Trainium."

"This was AWS updating the published base rates on their pricing page... $34.608/hr became $39.799/hr uniformly across every region. That's a policy decision, not supply/demand."

"The precedent is set. That's the part that matters. Once the door is open, it doesn't close. Every FinOps team just added a new line to their risk register."

AWS increased EC2 Capacity Blocks for ML pricing by about 15% across all regions where the service is available. The price rise affects organizations reserving dedicated GPU capacity for large-scale machine learning workloads, increasing rates on AWS's top ML instances including P5en, P5e, P5, P4d, Trn2, and Trn1. For example, a p5e.48xlarge with eight NVIDIA H100 GPUs rose to $39.80 per hour from $34.61, while a p5en.48xlarge with eight H200 GPUs rose to $41.60 from $36.18. Capacity Blocks reserve GPU instances within EC2 UltraClusters and guarantee access for defined periods. AWS has said pricing adjusts with supply and demand, but observers called the uniform base-rate update a policy decision that heightens FinOps risk.

Read at InfoQ