No company in the game to make a profit would sell ASIC hardware below the expected value of all Bitcoins the miner will produce during its useful life.
so in the days of the wild west no one sold a pickaxe for less the the average earnings of a gold miner in his life time
so in the industrials age no one sold a stick of dynamite for less the the average earnings of a gold miner in his life time
so in todays times no one is selling excavators and shloosh machines for less the the average earnings of a gold miner in his life time.
next you will be telling me that farmers have never sold a potato at wholesale to retailers but instead sold every potato they dug up at retail price...
The FABS would be the ones selling the pickaxe, dynamite, etc. The people contracting the fabs 'in-mass' to produce 1 million chips have two options, mine in-house, or sell to customers. Option A is 80% cheaper than option B. End result, margin on mining becomes so tight that the only people who can compete are those who follow Option A, in house mining. Option B has capital costs 4x as high before they can start mining.
Your analogies are poor because they are approximations because the intermediate products are MULTI-USE and have no value without a mine and labor behind it. Where as every ASIC cpu has value with only electricity as an input. Large mining farms will have much cheaper electricity, cheaper packaging, and economies of scale that will increase the difficulty to such a point that only people operating at such efficiencies can mine more bitcoin that the cost of the capital required to produce them.
. 
. It's hard to tell if they will ever pay off but I mean, if they last a few years I guess they could pay off eventually. 
