Preconfigured ML environments

Sale No Expires
Images come set up for common ML frameworks.
Get Deal
100% Success

Reserved capacity pricing

Sale No Expires
Commit to capacity for a lower effective rate.
Get Deal
100% Success

On-demand GPU instances

Sale No Expires
Spin up deep learning GPUs and pay by the hour.
Get Deal
100% Success

★★★★★ 5.0/5 based on 127 user reviews

Saving on Lambda

GPUs bill by the hour on demand, so the biggest saving is stopping instances the moment a job finishes rather than leaving them idle.

How to keep costs down

  1. Pick the smallest GPU that handles your model.
  2. Use reserved capacity for steady, long running training.
  3. Shut down instances as soon as a run completes.

Worth knowing

The pre configured images save real setup time, which is part of the value. Bigger GPUs cost a lot more per hour, so profile whether you actually need them before scaling up.