Great thread.
So I had a chat with a 40 year veteran of the US Electric Power Industry over the weekend and based on what he was telling me, getting power in the US for under 3 cents sustained is next to impossible. You can get it in shifts, during off peak, do peak shaving and other tricks of the trade but here is the bottom line, at least in the US:
Generators will make more selling into the 'market' and to the ISO's even if it means taking generation offline. The example was that if a generator can generate power at 1 cent per Kwh and a miner came to them with a 40MW requirement and said they would pay 2 cents (100% margin), the gerator woul turn it down because they can get 7 cents per Kwh in the market someplace else. There are a bunch of other regulations that add to the reasons why buying direct can't or won't happen anytime soon.
I asked why not build a power plant and plug in a data center/mining operation into it and get a direct feed - regulation and the fact that you would need to have two generating plants for redundancy and your costs double. If you had grid tie in or wanted grid tie in to sell excess or to provide back up load, the regulations REALLY kick in and are prohibitive.
So the notion of continuous cheap power is a notion, at least in the US from what I was told, again from a guy who remembers the first NERC and FERC meetings.
The second thing I just looked at was hash rate to difficulty ratio growth for the last 6 months and the hash rate has gone up 5.5 times in the past 6 months while the difficulty has risen 4.3 times according to the data on blockchain.info I looked at. So the hash rate is growing ahead of the difficulty still, which leads me to believe that is a good thing and money can still be made but you need to go big (decimegawatt scale, 10MW at a time) and throw as much hash rate while the difficulty slips... Right?
the idea in bold is wrong.
basically the current diff number is a 'real number'
any and all hashrate numbers are guesses and estimates.
the next diff number will be based on the time it took to do 2016 blocks (maybe 2116 as I had a beer or two tonight) so the diff number is always a true number and hashrate is a guess based on how fast you are making blocks.
Think of it like this. box cars or a pair of sixes with dice are a diff of 35 to 1 . So if I had 1 guy toss 1 pair of dice 10 times an hour it will take him a while to toss 'boxcars' or 6 + 6 . with normal luck it takes about 3.5 hours or 36 tosses to do it.
but if that guy tossed boxcars on ten tosses in a row. he would do 10 sets of ' boxcars ' in 1 hour .
it would make you think they were a few more guys tossing a few more pair of dice to get that many 6 + 6 tosses in only 1 hour .
your hashrate estimate would be very high