AWS Batch can split GPU capacity between teams for AI training
AWS Batch now lets teams split GPU capacity into quota shares for SageMaker Training jobs. Idle capacity can be lent to other teams automatically, and the highest-priority jobs can run first using preemption. It suits engineering teams that train AI models and share expensive GPUs.
How to try it
Open the AWS console, go to AWS Batch, and set up a quota share for one team to start.
Useful 35%