New study quantifies bitcoin’s ludicrous energy consumption

May 17, 2018 at 06:26AM
via Ars Technica

Enlarge / The bitcoin network uses several times the 1.21 gigawatts required to travel back in time.

Universal Pictures

The bitcoin network is run by miners, computers that maintain the shared transaction ledger called the blockchain. A new study estimates that this process consumes at least 2.6GW of power—almost as much electric power as Ireland consumes. This figure could rise to 7.7GW before the end of 2018—accounting for almost half a percent of the world's electricity consumption.

The study is an updated version of calculations performed late last year by analyst Alex de Vries. In this new version, de Vries has gathered more detailed information about the economics of the mining business. But his new numbers are broadly consistent with the old ones. Last December, he estimated that the bitcoin network was consuming roughly 32TWh annually, or 3.65GW. His website, which is updated daily, now shows the network consuming 67TWh annually, just under that upper bound of 7.7GW shown in his new study.

As de Vries makes clear in his new paper, these numbers are necessarily speculative. Bitcoin mining is a decentralized and secretive industry. We know how much computing power the bitcoin network has—right now it's about 30 trillion SHA-256 hashes per second. But miners are making these calculations on different types of hardware with different levels of energy efficiency, so we can't convert that figure directly to energy consumption.

If you assume that the entire network is using the most efficient known mining hardware—the Antminer S9 from Bitmain—that yields de Vries's lower bound energy consumption of 2.6GW. However, we know that people are using other, less efficient hardware, so the true energy consumption is probably significantly higher.

Economics provides an upper bound on bitcoin's energy consumption

How much higher? To establish his upper bound of 7.7GW, de Vries uses some economic reasoning. Bitcoin's rules allow the creator of a block to award itself 12.5 bitcoins—worth more than $100,000 at today's prices. With a new block being created every 10 minutes, that works out to around $15 million per day in mining revenues.

A crucial point here is that the difficulty of the mining task automatically adjusts to maintain a 10 minute average block creation rate. So if more computing power joins the network, the result isn't that more bitcoins get created. Instead, it takes more computing power to produce each bitcoin, making existing mining hardware less profitable than before—and driving up the energy consumed per bitcoin.

So de Vries calculated how much electricity bitcoin miners would have to consume for bitcoin mining to no longer be profitable. He assumed that electricity prices make up 60 percent of the cost of mining and that electricity costs an average of 5 cents per KWh.

That yielded energy consumption of 7.7GW, which serves as an upper bound of the bitcoin network's energy consumption. If energy consumption rose above that figure (assuming that the average electricity cost really is at least 5¢) that would mean that some miners would be losing money and would have to shut down.

However, it's quite possible that miners are consuming significantly less than 7.7GW. During times of rapidly rising bitcoin prices, like we saw in the fall of 2017, there might not be enough mining hardware to go around. In that case, miners would be generating bitcoins using less than 7.7GW, earning big profits in the process. People would be adding more hardware to the network as fast as possible, but it takes time for the makers of super-efficient bitcoin mining hardware to expand production.

What all this means is that the bitcoin network is consuming somewhere between 2.6GW and 7.7GW of power, but that we should expect energy consumption to rise over time until it reaches around 7.7GW. Given that bitcoin's value has been trending downward for the last five months, we should expect the price to be closer to the high end of this range. If bitcoin stays at its current level around $8,000, de Vries expects the network to reach 7.7GW of energy consumption by the end of the year.

If bitcoin's price were to regain its December highs of nearly $20,000, then mining would suddenly become hugely profitable again and should expect to see energy use rise much more quickly. On the other hand, if bitcoin's price crashed, miners would start turning off the least energy-efficient mining equipment, and energy use might actually fall.

To some extent, bitcoin's high energy use is a problem that will resolve itself over time. The bitcoin network is programmed to reduce the block reward by 50 percent every four years, with the next halving scheduled to happen in mid-2020. When that happens (assuming a constant bitcoin price) the mining industry's revenue will fall in half. In equilibrium that should mean that energy use falls by half as well. Energy use should halve again in 2024, 2028, and so forth—though that could be offset by further gains in bitcoin's price.

De Vries points out that the network consumes around 300kWh of electricity per bitcoin transaction, which is obviously a lot. But this figure is often misinterpreted. The zero-sum computational race that drives bitcoin's prodigious energy consumption has very little to do with the number of transactions per block. An empty block takes almost as much energy to mine as a full one. So getting people to make fewer bitcoin transactions wouldn't save significant energy; as a matter of arithmetic it would simply cause the energy per transaction to go up to even more absurd levels.