The Cortex-M3 core has 33000 gates.

In 0.13u geometry, the core takes 0.3mm2 of the wafer and draws 4.5mW (at 50MHz).

It's the fact that the transistors have shrunk so much in size that have made it so hard for 8-bit processors to keep up. The ARM dies are still very small despite all transistors - and their current consumption is still very small.

In the end, the chip manufacturers makes as advanced peripherials as they can, to get the customers to select their specific processor.

Much of the cost is in the testing - when they have good silicon it's just a question of dumping out processors in huge volumes.