It's basically because we have no solid-state way to reduce voltage without dissipating heat. When a solid state switch is fully on it conducts lots of current at almost no voltage drop, and when it is fully off it blocks full voltage at nearly zero current so losses are low.
Anything in between it behaves like a resistor. So to avoid lot's of heat and power loss we only use solid-state switches in full on or off in power applications. This rules out the simplest method of producing a sine wave for high power applications which requires a gradient of output voltages.
If you want to make a sine-wave while still using the switch only ever in full no or full off, then it becomes more complex than simple PWM because you are still PWMing to maintain efficiency in the solid state switches, but now you are adding on extra filtering afterwards to turn it into a sine wave.
If you can come up with a solid-state method to reduce energy coupling between two points while maintaining high efficiency then you can reduce voltage without incurring losses, similar to how the coils move past each other in a generator and overlap to varying degrees producing different amounts of coupling. If you figure that out, then you can produce a graduated output without incurring losses much the same way a spinning generator is able to make a sine wave without incurring losses. Then we will have efficient, simple, straightforward high power sine PWM. You will also probably win a Nobel prize.
