Spinning Straw into Gold: The Economics of Bitcoin Mining
I’ve been reading up a little on the supply-side of Bitcoins, and it’s actually very interesting. Currently the estimated electricity cost per Bitcoin mined is approximately $35 (data from here), and it’s still trading at nearly $100. So it’s still a profitable occupation - if requiring the purchase and use of a lot of increasingly complex, expensive hardware - but the intriguing thing is that the system is self-correcting, to an extent, between the level of demand for Bitcoins and the ease of supplying new ones.
The chart above shows profit margins based on estimated electricity costs and actual Bitcoin prices; according to it, for nearly two months around the start of the year miners were operating at a loss (in December 2012 the ‘block reward’ halved, reducing the number of Bitcoins created by each mined block - hence the sharp drop in profitability). The market price (demand side) didn’t change much, in fact it kept climbing past the $13 mark; while processing activity only dipped slightly, despite producing revenue of only half as many Bitcoins. Indeed, the fact that anyone continued mining at all indicates the strength of confidence in Bitcoin’s future appreciation in value (as well as the halving of the demand reward being an integral feature of the system, known in advance - the next one is due in 2017).
What regulates the supply of new Bitcoins is the difficulty of obtaining them, which is adjusted:
"every 2016 blocks based on the time it took to find the previous 2016 blocks. At the desired rate of one block each 10 minutes, 2016 blocks would take exactly two weeks to find. If the previous 2016 blocks took more than two weeks to find, the difficulty is reduced. If they took less than two weeks, the difficulty is increased. The change in difficulty is in proportion to the amount of time over or under two weeks the previous 2016 blocks took to find"
In other words, it adjusts to the computational power of the network (i.e., of all Bitcoin miners in operation). On this chart it can be seen that, following the drop in activity after the halving of the block reward in December, it decreased in January before rising again in March and climbing sharply thereafter with the increased interest in, and demand for, Bitcoins. As Bitcoins are worth more and so more people mine them using more powerful equipment, the difficulty increases in proportion to their success so that roughly the same amount are created each month (approximately 100,000) regardless of the resources being put in to find them. Which on the one hand is really very clever, and on the other seems incredibly wasteful: as more people mine Bitcoins, the more the difficulty goes up, and the more power and equipment has to be put in for anyone to compete, without increasing the long-term efficiency of the process.
It’s like a combination of Sisyphus and Tantalus: always running to keep still, or reaching for a retreating target. Of course that’s thinking of Bitcoins in purely nominal terms; with the increase in price, the demand incentive has more than kept pace with the increase in difficulty and mining supply, as shown in the rising profit margins in the chart above. Conversely, if the price continues to decline the rate of supply of Bitcoins may decrease, in which case the difficulty and thus the cost of mining them will fall. So while on first glance it may seem as though current miners are reliant on a high price for profitability, in fact the cost is tied to the difficulty; which is in turned tied to the volume of mining.
Think of it as a triangle between economic value (say, in $USD), Bitcoin (BCT) and processing power (H/s): the connection between the first and the second is the price, and the second and the third difficulty, and the third and the first cost. Price is largely independent, if one assumes that the addition of 100,000 BCT per month to a market of some 11 million is relatively marginal; but difficulty and cost are interdependent, both on each other and on price, mediated through the activity of Bitcoin miners. A (dis)equilibrium is thus maintained where the level of difficulty is such that the cost of mining is less than the given or expected price, but also such that the number of blocks mined don’t greatly exceed the fixed target.
The costs factored into the above charts, however, aside from being estimates based on varying electricity prices and power consumptions, are only the marginal cost per calculation: that is, they don’t include the cost of hardware (or indeed associated overheads)*. Already the competitive nature of Bitcoin mining has produced an ‘arms race’ in processing power: early coins were mined on CPUs, then GPUs (graphics cards, which were found to be able to perform the necessary calculations faster), and something called Field Programmable Gate Arrays (FPGAs), and now ASICs - Application Specific Integrated Circuits, or as the name suggests processors specifically designed to perform Bitcoin mining. Given the level of difficulty at current prices/costs, it becomes uneconomic to mine without the more powerful and power-efficient systems: yet if difficulty and/or price fell (as a cause, or effect, of miners leaving the network or reducing their activity) then the purchasers would be left with hardware that is impossible to use to its full potential.
However, if the price of Bitcoins rises again, the pressure will increase to maximise processing power. Since by the nature of arms races and the limit on Bitcoin production, this will do nothing but maintain miners’ place in the competition. One would wonder if it could reach the stage of providing a significant stimulus to processor speed development, even if the use is too specialised to have any direct benefits to computing generally - or, perhaps more importantly, to energy-efficient systems. The hard limit to such an effect will be the economic value of Bitcoin, although in the broad sense of the word - as the chart above shows, expectations of its future value are enough to pull in considerable resources. If Bitcoin continues until 2017, it would be interesting to see what effect the next halving of the block reward (and consequent doubling of mining costs) has on the supply.
The overall architecture of the Bitcoin supply is an exponential curve in which the remaining coins are increasingly scarce, but the difficulty of finding them varies according to the effort expended. Which is quite a clever, almost elegant, combination - it will respond both to the predicted increase in processing power and to the unpredictable variation in demand. Economically, though, I’m still not convinced it makes very much sense: for a currency that is aimed at eventually becoming an inflation-proof ultra-gold-standard, it seems odd to have the equivalent of very slow and steady quantitative easing preceding that fact. Inflation, in other words, that occurs at a fixed rate irrespective of whether the use of the currency is booming or not. Probably the rate is so weak, at around 4,000 BTC per day or 1% of the total volume per month, that it makes no impact on the inherent deflationary effects (hence my assumption that price remains independent of the marginal increase in supply) running up to the bubble, nor exacerbates the glut of supply afterwards - although it may assist with liquidity.
The operating profitability of the Bitcoin mining network closely tracked the bubble value of Bitcoin, reaching a peak of 85% and falling to a low of 57%. With the recovery of prices to around the $100 mark, it has risen again to 68% - about the same as it was when the price was at that level too. So clearly within that time period Bitcoin miners made greater profits but faced no greater competition, whereas over a longer period of sustained high prices you’d expect profit margins to reduce as miners eroded their own advantages through pushing the network towards increased levels of difficulty (the alternative would be a cartel where miners agreed a set rate of computation in order to keep the difficulty down; this wouldn’t, unlike normal cartels, affect the ‘consumer’ of the supply, but it would be antithetical to - and thoroughly endanger - the decentralised nature of the network). What would be left of profits would be what is necessary for investment and a reasonable return on the risk of staying in an uncertain game.
So while I think the supply side of Bitcoin is fascinating, and somewhat detached from the economic oddity of its demand side as a currency or commodity, it’s still a thoroughly neoliberal project where the accumulation of wealth is its own end; economic activity is directed towards out-competing others, rather than achieving tangible social or even technological goals; and individuals leverage capital for massive profits. Miners won’t destroy Bitcoin, but they remain reliant on the confidence of the currency’s holders for its value. You’ve gotta admire the computer nerds for finding a way to turn their hardware into a literal money-spinner - at least as long as it doesn’t all evaporate by morning (or they’ve banked some of their profits into hard currency).
*the Blockchain.info stats page also gives a lower “profit margin” of just under 50%, as against a 68% “operating margin”, for which “hardware costs are estimated to be $1000 per gigahash every 2 years, and bandwidth $1 per gigahash per year.”.