AWS Lambda Managed Instances

Allow me to introduce AWS Lambda Managed Instances. You can now choose what underlying hardware your Lambda functions run on for better performance and cost optimization.

Unfortunately, I haven’t found anything about running them on GPU instances yet, so that might not be available quite yet.

Interestingly enough, you get charged for 3 things: $0.20 per 1M invocations, the EC2 instance’s normal compute hour costs, and an additional 15% of the EC2 instance cost as a “Compute Management" fee.

I’m curious how this weighs out over just using Lambda. I suppose if you had a fairly constant and predictable amount of invocations, it could weigh out in the end. 

Question for you:

Do you have a use case for AWS Lambda Managed Instances?