The distinction between "microcontroller" and "microprocessor" is not a very helpful one. The way things are generally marketed:
Desktop/server CPUs: performance orientated. Not capable of driving their own peripherals; this is usually done by an associated "chipset". For this market, performance is most important, so new devices are developed every few months and fabricated on the latest factory lines. This greatly increases the price. There's not so much an "ALU" as a deep, broad pipeline of computing elements, and often quite a lot of the die is cache. An Intel CPU will have more RAM on it in the form of cache than any "microcontroller". Contain hundreds of millions to (NVIDIA GPUs) billions of transistors.
System-on-a-chip: moderate performance with most peripherals on board. Possibly RAM stacked on top in the same package (not same die). Target market is phones, tablets, mini-PCs, set top boxes. Price/power/performance are balanced against each other. Still fairly cutting-edge.
Microcontrollers: not required to do very much computation. Will generally spend their entire life running one application. Peripheral-orientated. Very cost-sensitive and maybe power consumption as well. The peripherals are usually not expensive, either in terms of design work or die area - they're pretty simple and don't develop over time. Although some of the fancier ones may have effort spent on high-quality or high-speed ADCs. Not manufactured on cutting-edge processes, which keeps cost down. The product lifecycle is a lot longer - some customers won't buy unless they're guaranteed a supply for a decade or more.
Midrange "peripherals" like USB and Ethernet support are complex, but can be bought in by the designers.
Fancy high-temperature or radiation-hardened microcontrollers are available, at extreme cost.