Shouldn't a watch last more then 3 years (especially a watch that cost hundreds of dollars)?
Theoretically yes; in practice though, that's rarely the case for first-generation anything that is reasonably complicated and/or groundbreaking, for a number of reasons. Was the original iPhone all that hot after three years? No, it was not.
Why? Well partly that is because you can't know in advance what exactly you'll need, so you build what you think you'll need, using whatever bits you have available to you, which won't be ideal ones because nobody's built that kind of device before. Also, you don't have infinite development time to make something exactly the way you want it. Original iPhone used an off-the-shelf ARM SoC that was admittedly pretty terrible, but as modern smartphones didn't exist yet there was nothing more suited available.
In the years to come, iPhone performance would explode each year, increasing much faster than the computer industry as a whole because suddenly there was a need for very high performance processors running on battery power, something that hadn't existed previously.
It's basically the same thing with Apple Watch. It needs to do roughly the same amount of work as an iPhone, on a MUCH smaller battery, and designing a processor for such a task takes years from start to finish; the S1 system module from original Apple Watch used the same SoC as a variation of iPhone 5's SoC IIRC, with only one CPU core in it (develped for one of the earlier gens of Apple TV I seem to recall), as Apple didn't yet have a chip that was entirely custom designed for a watch.
Now, armed with much more experience in building watches, including custom tailored processors and support chips (like the W2 and so on), that's already started to pay off bigly. With smartphones we saw this in iPhone 4, and I hope there will be a similar significant step up in performance and capability with the next Watch.
