Have a look at the drawing below that shows three different classes of analog ICs. At the top are very specialized devices that use up lots of company resources. If you are the guy making this part, you probably have only one such device in your portfolio and you almost certainly sell to only one customer.
For that ASSP at the top (shaded purple), you probably have a commitment from that customer to buy at least a million units over a reasonable time frame (e.g., one year). That is offset by what is certainly a large NRE. Could be profitable; could turn into a money pit if you have to re-spin the silicon too many times.
If you are supplying a somewhat more general purpose device such as an AFE (shaded green), you probably can find a lot more customers. These customers will be able to use such a part because it has fairly conventional devices contained within: op-amps, tightly matched resistors, voltage references, delta-sigma ADCs, etc. The typical customer can adapt this part to their design (or design concept) relatively easily. With a couple dozen customers, your risk is much lower. With a device that can be adapted to several different sorts of products, customer acceptance is increased — again, lowering your risk.
And if you're supplying commodity parts (op-amps, voltage references, DCPs, LDOs — shaded yellow), you will have plenty of potential customers, but also plenty of competition and very thin profit margins.
While I am at DesignCon, I am visiting with IC manufactures and discussing these sorts of devices and these sorts of marketing issues. These thoughts are based on those discussions. Let me know your own thoughts on these issues.