Amazon Launches New Service for Reserved GPU Access to Boost AI Model Training

AWS launched Amazon Elastic Compute Cloud (EC2) Capacity Blocks for ML, letting customers rent Nvidia GPUs for a defined time period for AI projects.

The product provides access to NVIDIA H100 Tensor Core GPU cluster instances that customers can reserve up to 8 weeks in advance.

It enables customers to reserve the GPU compute capacity they need for the exact time required to run AI jobs like training models.

Customers get cost certainty upfront by reserving capacity like booking a hotel room for specified dates.

The price for access varies dynamically based on supply and demand when users reserve capacity.