The last mile is very expensive. It costs a ton to maintain that infrastructure.. Yeah, so I can add a downstream DOCSIS QAM to an existing chassis for $50.. Say 8*50 = $400, it is just the beginning. Cable modem termination systems? Go buy and maintain a bunch of uBR10012s and ASR routers and tell me how cheap it is. The taps, the wiring, the amplifiers, the trouble calls cause the plant is old. Do I need additional chassis to handle, will the lasers even do it, if I split the node how far will I need to trench new fiber? Do I even have spare fiber back to my headend? Will it even fit in my existing building, do I need more AC? Have I maxxed out my power panel?
The broadband CEOs quoted have no clue.. They clearly don't work for tier-1 MSOs.. Multiply by 10s of thousands and it starts getting real -- very fast. Anyone can run 50 service groups out of one small town. (E.g. Sonic?) Frontier hardly faces the same constraints.. They have distributed DSLAMs and offer crap speed through them.
In the real world, broadband demand is quickly outstripping cable plant capacity. Just adding 8 or 16 QAMs would be too easy. (Assuming you have it and a lot of plants don't.) This means - more money. Plant upgrades to 1.2 Ghz, CCAP, all active/passive component upgrades, and it still only buys you a few years the way things are going.
If it were up to me, municipalities would run fiber to each house and then connect us to the provider of our choosing. It's a nice dream.