I was very sceptical about data in the cloud, mainly because the early software implementations I saw were very much based on functional requirements with scant concern for non-functional requirements such as security.
Having spent time with one of AWS's architects with a specialism in security the one thing that doesn't worry me is AWS security. That's not to say I don't worry about security in AWS but it is not the vendor's side of things that concerns me.
Where I see a gap is in the education piece around cloud security. The vendor documentation is good and what you need is included in other documentation. The bit I want is to address my paranoia. Highly specific, Security for Dummies "This is how you set up an RDS instance so only people in your organisation can access it".
IBM withdrew Netezza because the market for big tin is evaporating. DWaaS makes a lot of sense provided the security requirements are satisfied. Snowflake and Vertica EON make a lot of sense. These are column store databases with S3 buckets providing the storage and incredible resilience. The actual compute spins up as and when you need it independently of the storage. This can make for far more intelligent cost control decisions. If a large, but short term analysis job is required then OpEx provides agility rather than CapEx.