Yeah it's our mistake. It's actually gigabyte-hour.

So in the OP's example, the cost is $87,000 per petabyte-year of storage?

EDIT: I see mentioned in another post this a per-replica cost. So it would be roughly $87,000 times the number of replicas, ignoring the initial queries that inserted the data in the first place?

I think so but we are going to run the cost regressions again and make sure it's in line with market.