SQL Query or Remove Columns

Yes and no, and maybe depending on what your next steps are :)

From the basic example you listed, you're getting the exact same result, and it wouldn't really matter one way or another. However, when you use Get Data to pull from certain sources, Power Query uses something called query folding. If you import the whole table from SQL, then remove everything but the two columns in the Query Editor, Power Query is smart enough to realize what you've done, and essentially write back your SQL to only pull those two columns. The first load to Preview might take a while since you started with all 100 columns, but refreshes after that will know you only wanted two columns and will grab those, pushing that "smarter" request back to the database.

Filtering is your WHERE clause. You could filter Column A to only include values of XYZ, and again, query folding will push that back to the database to do the work instead.

However, certain things will break query folding and prevent it from pushing the heavy lifting of your query back to the database. For example, writing custom SQL automatically prevents query folding from taking place. If you used SQL to specify those two columns, then used the query editor to filter Column A for XYZ values, that filter can't be pushed back to the database as a WHERE clause.

Which should you use? If you know SQL well, and you know you can get your data the way you want it without further transformation (or at least without further transformation that would prevent query folding anyway), you might as well use SQL. But if you think Power Query could write the SQL better for you, just pull in the whole table and do all of your filtering, transformations, etc., from there. Power Query will use query folding as much as possible to help you out and push the processing back on the database.

If you want to know more, here's a pretty good breakdown of query folding to get you started. Fair warning, it's not been updated since May 2015, so there may have been changes or improvements since the time of the article.