As Aberforth mentioned, since the radius of the platters is fixed by the physical size of the drive, there are two variables that can be used to increase the size: density of the bits and number of platters. Obviously, the number of platters will also become limited by the mechanical size of the drive's enclosure, so this sets an upward limit on the number of platters. The density of bits on the platters is set by a few different, interacting considerations: the ability to process the platter with bits at the specified density, electromagnetic interactions between the bits as they become very close together, and the ability to control the read head when trying to differentiate between bits. All of these things will influence performance to varying degrees. For example, if the bits are closer together, the head doesn't have to travel as far to read consecutive bits. Improved control drops the time required to find the bit you're looking for.
There have been a lot of patents recently on the motor/control side of things recently, and a lot of work on the processing end. I'm not sure how close the technology is to being limited by electromagnetic interactions at this point.