AWS has launched a service that allows customers to rent NVIDIA GPUs for short-term AI projects

Cloud provider AWS has introduced Amazon EC2 Capacity Blocks for ML. It allows customers to rent NVIDIA GPUs for a specified period of time to run AI projects.

Here's What We Know

The service provides access to NVIDIA H100 Tensor Core GPUs. Customers can book from 1 to 64 units for a period of 1 to 14 days. Once the period expires, instances are automatically deactivated.

This provides the ability to accurately calculate the required computing power and budget for the project. The price for access is dynamic and depends on demand.

Before purchasing, the customer can adjust the number of resources and the lease term after seeing the total cost. In this way, the costs of AI projects are optimised.

The service is already available in the eastern region of AWS US (Ohio).

Source: Amazon