This gen of asic might run longer than the last two? People seem to be building out their farms for the 12.5 btc era right now. Some of the older gear will turn off for the halving but it will be outpaced by new gear coming on. Ramping the difficulty all at once has a habit of deterring competition (especially from the small guys), which means greater profitability during the lifetime of the hardware (optimistically 3 years).
Expect more of this stuff:
https://megabigpower.com/indexProvide financing to fill out a current warehouse, or if you have a space, they'll provide and install the hardware, you split revenue 50:50.
Short term things are fine, long term's a little more dicey. I see 4 major options.
1. Block size and number of tx raises, continuing the low cost value transfer attribute of today. Miners chose appropriate block sizes to maximize profit along their demand and supply curves.
2. Relatively the same amount of hodlers as exist today are fine with $3-7 fees per tx. Former payment users are (partially?) replaced by digital-goldbugs. Bitcoin stays around the same size in terms of market cap, maybe grow some or shrink some, nothing fatal.
3. Payment channels emerge from vapor stage, the added complexity and need of opening channels by locking up some btc isn't an impediment to network growth... Large fees are paid when these scripted arrangements are settled to the blockchain. Although LN is open source and free... Blockstream is added to the s&p 500 somehow.
4. (3.) doesn't really work. Former users of btc migrate to competing options, some hodlers tuck away the private keys and hope for the best, some muck (or have mucked) their hands and move(d) on.