Our systems are predominantly cloud native. For data warehousing handing someone else the headache of exponential data growth is extremely attractive.
Our infrastructure manager said that one of the attraction of the cloud is that you can scale down. You no longer have to try and guess workload increase over the lifetime of a server you buy for your data centre. You rent for the demand that you have today and if demand materialises, then you choose the next instance size up. That is, if your chosen tech even has an instance size.
Google BigQuery charges you on the basis of the data volume you query. Physical storage and compute power are managed for you.
AWS Lambda functions and Google Cloud functions spin up and execute when they are called. They can save a lot of money if the calls to them naturally ebb and flow or are functions called infrequently.
The range of data science tools available in the cloud is greater than a company would have if buying for on premises installation. That is the hidden cost of an RFI/RFP process largely dispensed with.
DBaaS is attractive to many because their DB workloads are the sort where the on premises equivalent would be one the DBA would watch but probably rarely need to touch.
That said, old school DBA skills do come in handy diagnosing performance issues though the nature of a managed service is that many of the tweaks to configuration are not available to you. You have to design around the problem. Getting the data model right becomes the most powerful weapon in your arsenal.