Land leases are expensive for tower placement, and they need electricity and a backup generator, to power the comm. equipment and the air conditioning.
Don't the utility companies have the same issues with infrastructure placement and land though? Or is there some eminent domain thing for power utilities that doesn't apply to telecom?
Well, in the US, electric and telephone have more rights and regulations, compared to cable television.
So, as a telecom, it is better to go in under the guise of a CATV company and use a franchise agreement, then as a phone company that also offers TV, and internet.
Also, if the poles are already there (usually phone or power), the new entrant is able to lease a portion of the pole to string infrastructure.
Sometimes the municipality owns the power company.
Cellular providers lease tower space from private companies that negotiate the tower placement with private land owners.
Specializing in hardware design is really a poison pill despite how interesting the field is. You spend more than a decade in school to gain the skills needed to work at the chip level, which often requires a PhD. Then it turns out that there's only a handful of companies in the world with the equipment or supporting personnel needed for the field, and your skills are worthless outside of these specialized environments. No wonder pay is so low despite the years of schooling needed. After going through those considerations, I ended up dropping out of a PhD in RFIC design and fell back to embedded systems. More companies hiring and versatile enough to slide into general software or web programming if necessary. Chip design is still fascinating, but all traps look nice on the outside.
Also, for any aspiring undergrads wandering this way that thought RFIC design was fascinating because they picked up a copy of Razavi's "RF Microelectronics" or something similar and couldn't help but read it cover from cover, the textbooks lie. The analytical equations in a chip design textbook are not reflective of what the actual process is like. While you're still in college, try and get a professor to give you a student addition of Keysight ADS or Microwave Office AWR, along with an SDK from a fab. Then spend a few months of simulation time to see for yourself how much simulated results diverge from analytical models and how painstakingly slow it is to optimize those simulations especially once you start moving into EM simulations. And keep in mind these are simulations, chances are the physical chip will diverge even more due to process variations, unknown parasitics, or even bad simulation models from the fab. If the lack of precision, simulation time, and general drudgery doesn't kill your enthusiasm when compared to your textbook learning, then by all means proceed.
It's still a fascinating subject. Every so often when I need a break from code I'll take out my old RFIC textbooks and read a chapter or two. But you really don't want to know how this sausage is actually made.
I spent my teens looking up to microelectronics career, and actually self-studied a year, or two ahead for entrance to Nanyang, or NTU semi.eng. programs.
I almost accidentally met two ex-TSMC engineers who worked for Chang almost since his first days at ITRI, and they were very good mentors, and seriously warned me about "a decade+ grind ahead." In the end it were my parents who stonewalled the idea, and sent me to a business degree mill in Canada. And this is all was when I had 2 early admission letters from both Nanyang, and NTU on my desk...
> If the lack of precision, simulation time, and general drudgery doesn't kill your enthusiasm when compared to your textbook learning, then by all means proceed.
OK, so 100k at most per engineer for "over 100" engineers, and those are just the ones that jumped ship, never mind the thousands staying behind. Is that significant?
Considering the average TSMC engineer makes 50k-60k, a 100k increase would probably be taken quite well.
Taiwan has such a big surplus of engineers and physicist that they don't really make much. Unless you are part of the club that knows TSMC special sauce in which you will be ridiculously well paid even by American standards, but that's not a large group.
Also, fiber can be analog, too.