It's my former company now, but the answer is: For the initial workload that we migrated from the datacenter to AWS EC2 instances, we still had the existing DevOps people, and while they said things were a little easier, they were still pretty busy.

For all new dev, which eventually including migrated some functionality from that initial workload, we used AWS-specific managed services like Lambda, Dynamo, and SNS/SQS, which required MUCH less effort from DevOps. Two of the existing DevOps people were able to handle everything the so-called serverless development required in addition to their previous normal work.

I mean, the point of managed services is that the vendor takes care of managing them, so you don't need to deal with upgrades, security patches, etc, so if you're still doing just as much DevOps work, it's likely you're not taking good advantage of managed services!

So we had some initial effort to automate and standardize CloudFormation configuration, and then developers were able to launch new code bringing in millions in revenue with very minimal oversight from DevOps.

The advantage there, though, was mostly from "serverless" and managed services, not just EC2 vs datacenter. We still had a DBA for the Postgres stuff, we just didn't need one for the new Dynamo stuff.