It's only partially a question of quantity. The larger factor is the amount of labor involved in dealing with a bad drive. Their (large) operation seems autom

It's only partially a question of quantity. The larger factor is the amount of labor involved in dealing with a bad drive. Their (large) operation seems automated so I assume if they have a single drive failure in machine #1234 (which has 50+ drives) they have an automated way of switching off said drive. Then they leave it there until the entire rack is replaced many years later. Other operations have to send a human to manually replace every single failed drive as soon as it fails, which is very expensive as a % of the cost of the bare drive.