Because most of those drives are getting slotted into sleds in arrays. A double height drive might work fine in a PC case (at least if you're willing to break out a dremel), but the majority of the buyers of big drives are not using ordinary PC cases.
Most of the bandwidth of a drive is out at the edge of the platter, so you don’t want to make them smaller in order to improve the seek time or make the read arms thinner to stack more platters. Which increases capacity less dramatically each time anyway.
There’s an old trick though that if you want a faster drive, only partition part of the space on each drive and thus reduce the average seek time. What one could do is make the axle bigger, shorten the arms, make them a little thinner and add another platter. But the next one is only 9% and that’s if you don’t take 9% off the area to accomplish it. Otherwise you get the same space but faster seek time.
