Agreed GPU miners in Europe other than those where someone else was paying the power bill ("free power") were the first to drop out. Right now hardware costs dominate but over the next year prices will drop, mining margins will be compressed and just like GPU mining power will make up a larger portion of the bill.
One also has to consider the geopolitical risk as well. Ask companies how much they lost when Chavez started nationalizing foreign assets in Venezuela.
You can't do that for a PH/s or more. Take the price of the business property lease, add in employees, add in security (you aren't going to leave you $50M in miners alone at night are you), add in insurance cost, add in fire suppression (might be mandatory in some areas), throw in business taxes, and pile on the additional energy and capital costs for 800 KW of industrial chillers, etc. Even mundane stuff like enough racks to mount 1 PH/s (you aren't going to just leave them on the floor are you, and the cost of an electrician to run enough drops to power 1,000 or more systems all add up. All those are Diseconomies of scale. The company may be cheaper but GH/s but it isn't 10x cheaper like the raw power costs might suggest. The last thing to consider is cost of capital. After this gold rush period is over mining isn't a particularly attractive venture from a risk vs reward point of view. Many investors may look at the small reward, the outsized risk, and the fact that profits are largely determined by the actions of outsiders and simply pass on the millions necessary to build up such a mining farm.
One thing I am particularly interested in is using the "waste heat". For a large operation (say 1 PH/s = 800 KW = 1 million BTU/hr) they have a lot of waste heat but it has no value. It actually has negative value as the company will spend money both on hardware and energy to remove that heat. On the other hand a home user could use waste heat to preheat hot water (free hot water for the rest of your life), and in the winter supplement heating the residence. With GPU rigs they are too bulky, complex, and skill intensive. However take some ASICS put them in a compact sealed box, cold water in, hot water out and that "waste heat" suddenly has value. I am doing some experiments right now. Interesting stuff.
That being said I do believe large commercial players will make up a significant portion of the network capacity in the future. I still think hobbyists and small home farmers will exist however they will be in areas of the world with lower than average energy prices (<$0.10 per kWh), generally cooler climates, and may look to thinks like using waste heat to boost ROI%.