Globally redundant and available database by accident?

Given the recent S3 Lambda service — https://www.zdnet.com/google-amp/article/aws-rolls-out-s3-object-lambda-to-process-data-for-multiple-applications/

If S3 is truly global (multi az-zones), redundancy, and infinite storage, instead of using Cassandra — why not save customer data as text files and use lambda to convert to Json for each application?