It's not arbitrary but nor is it very scientific. It's an informed guess from an interested punter. The last time I heard figures, it cost around $3300 in elec

It's not arbitrary but nor is it very scientific. It's an informed guess from an interested punter. The last time I heard figures, it cost around $3300 in electricity for a given mining farm to mine a bitcoin on average. That cost will go up and down based on the number of people mining but in my observation the cost of mining is usually pretty close to the value of a bitcoin. I can't prove that that isn't coincidence, but to me it makes sense. If the value of bitcoins goes up, people will online mining farms as it becomes profitable in areas with more expensive electricity, and the difficulty will go up. If the price drops, people offline their mining farms. On the other hand, the approximate cost of mining a bitcoin is public knowledge since the current block difficulty is public knowledge. Anyone buying large amounts of bitcoin can make very accurate estimates of a miners profit margins with nothing more than the miners IP address and the block difficulty, which puts miners on the back foot for negotiation. When it comes to the profit margin of miners, one component is the electricity price, but the other is amortizing the cost of the mining hardware. Graphics cards are relatively cheap ($200-800 depending), but ASICs can go for upwards of $12000-$15000. As the number of mining farms goes up, the pressure on mining hardware manufacturers to maintain a steady supply of mining hardware increases. If they cannot do so, the cost of replacement mining hardware increases and starts to eat into profits. If the supply gets too short, the cost of replacing the hardware can become so high that the real profit post-amortization is flagrantly not worth the effort, particularly in places with expensive electricity. Miners who are in it for the money usually want to see their profit completely amortize the cost of their hardware within three months of buying it. If it takes longer than that, you're straight up gambling that the current price won't drop before you can make a profit. GPU miners also often replace their cards on a 3 month cycle and sell the second hand ones on for gamers to use, which helps with the amortization situation. My spitball estimate is that if ASICs remain dominant, $4000/BTC is the equilibrium point of it being profitable in some places but not others and the number of miners balancing the supply and cost of ASICs such that ASICs can be replaced at a steady rate without supply problems and still pay for themselves before they burn out or become obsolete due to a new generation of ASICs being developed. $400 is the same spitball estimate, but assuming that ASICs fall out of favor, the difficulty drops, and all those etherium GPU rigs start moonlighting with bitcoin on the side.