The Tableau Data Engine can certainly handle extracts of this size, and depending on the type of data you may find that Tableau performs better with Data Engine extracts than it does with the underlying database. However I'll point out that Teradata installations are typically very fast (though you can buy archive-oriented warehouses which don't perform as well). You will likely find that working with our first-class connector to Teradata yields such a good experience that you won't typically need to create extracts.
Thanks, Robert. I don't really see you coming down firmly on one side or the other, but it's good to know that both methods can be effective -- so we will probably stick with the Teradata connection. I appreciate your help!