What's the basis of your statement? For data warehouse/data lake type applications, data typically goes in and aggregates come out. The point of these systems is to fish out only the data you actually need to make decisions, which is usually much smaller than the source dataset. Compute and storage therefore tend to dwarf other costs.