Data Engineering
ETL and ELT
The transform step moved. That single change is what the warehouse era is actually about.
Grasp
Both spell out extract, transform and load. The only difference is the order, and the order is a statement about where the compute lives. In ETL the data is reshaped on the way in, by a machine you run, and the warehouse receives it already clean. In ELT the raw data lands first and the reshaping happens inside the warehouse, in SQL.
The industry moved because warehouse compute became elastic and separable from storage. When the warehouse can scale to the transformation, a separate transformation server is a machine to size, patch and pay for while it idles, with no benefit. Landing raw first also means a bug in the logic costs a rerun rather than a re-extraction, because the source data is still there. That is why the transformation layer turned into version-controlled SQL that can be tested and reviewed like any other code.
The trade is worth saying plainly, because "ELT is modern" is not an argument. You have moved the bill onto the warehouse, which charges for every scan the transformations make, and you have landed raw data that may contain personal information you are now storing and must be able to delete. Neither is a reason to avoid ELT. Both are reasons to know which you chose, and why.