[edit} OOPS, mindlessly I have been making statements thinking FPGA, the cost differentials below are way more re ASICs. In ASIC volumes it is even sillier to go for "universlity". I think Richards estimate of $2 million is a bit high, I think one million will do. The trend these days is moving towards FPGSAs simply because of that cost. Attila, does you ASIC designer pay the added half mil if there is a small timing error or such?[end edit]

In the reality we plan to use ASIC for more than 10 different applications, and one of the application is where we for sure don't need UART. And from these produts we plan to sell millions of pieces.
You keep on about not needing the UART in ONE of the apps, That must mean that you need it in the remainding 9. So, to save $0.10 in one app, you are willing to add $2 to 9 other apps - that would not make sense even to a two year old.
Use DEDICATED ASICS the idea of an ASIC is to match one project perfectly and you should FORGET about "it would be nice if we could use it for other things". If you sell "millions of pieces" the cost of developing an ASIC will "disappear". If you want something "universal" that is what the uC makers sell and not an ASIC app.

I have never heard about any "universality" (ASIC PC board or whatever) wanted from anything above, say, 10.000 units.

Erik