I think this is all very chickens**t. That being said, I would suggest you look at data retention as a potential revenue stream rather that an unrecoverable cost that must be arbitrarily capped.
You could offer paid data retention options for those who need it. Something like a free base amount (10 gb?) plus perhaps $10 per year per additional 10 GB blocks (or $100 per year per 100 GB blocks). With the data retention option, you should offer the user full control over their data, including ability to freely download and/or archive their own data. Give the user ownership and control over their data and the ability to maintain their own balance between data retention and cost.
