It’s no secret that Bitcoin (BTC) mining is an expensive business, and in more ways than one. Not only has it become less profitable since July 2016’s halving of mining rewards to 12.5 BTC, but competition among miners and an increasing hashrate have resulted in ever-higher energy consumption, with all the damage to the environment that implies.
Yet, as energy-intensive as Bitcoin mining is, a question still remains: Is there a seasonal variation in the cryptocurrency’s energy consumption? Even if consumption is rising on the whole, does something different happen during the summer months?
Well, data hasn’t been collected on Bitcoin’s electricity consumption for long enough to provide a truly authoritative answer to this question, yet what data there is suggests that the summer brings a slight, but noticeable weakening to the rise in BTC’s energy consumption. This is most likely because, globally, energy prices increase during the summer months, putting a strain on the profitability of Bitcoin mining.
Steady growth
When it comes to the question of Bitcoin’s energy consumption, the first thing that needs to be stated is that direct data on consumption hasn’t been made available by the big mining companies. Still, a number of indirect estimations have been produced over the years — based on such metrics as profits, network difficulty and hardware efficiency — and these all show that consumption has been increasing consistently.
Back in June 2014, the first rigorous study on BTC energy consumption was published by Karl J. O’Dwyer and David Malone of the National University of Ireland Maynooth. It estimated Bitcoin’s annual energy cost to be something between 0.1–10GW (accounting for the uncertainty as to which mining equipment was being used), although the authors settled — though without fully explaining why — on 3GW, which was equivalent to Ireland’s yearly consumption level at the time.
Since then, the most widely cited data has come from the Bitcoin Energy Consumption Index (BECI). Produced by analyst Alex de Vries, the BECI settled on a higher figure than that of O’Dwyer and Malone’s model, and it has continued to reveal steady, day-to-day increases in BTC consumption ever since it started collecting data in February 2017. In December, it put annual consumption at 32TW/h per year — equal to 3.65GW. By contrast, its latest figure — for Sept. 12 — indicates that the Bitcoin network is now eating up 73TW/h — or around 8.8GW – each year. However, in a standalone, peer-reviewed paper from May, de Vries put annual consumption at 2.55GW (22.4TW/h).
As the table below illustrates, de Vries’ data shows that there have been very few dips during this overall rise. The strong increases continued even during the first half of 2018, when the BTC price saw a considerable correction from its December high of $19,900. For instance, when the price fell by 46.2 percent over three months to the Feb. 17 price of $10,707, BTC’s energy consumption increased by 42.6 percent over the same period — from 34.96TW/h to 49.85TW/h. And when BTC’s value dropped by 9.87 percent between April and the end of June (to $6,366), its energy consumption rose by 20.9 percent (to 71.1TW/h).
This goes to show that, despite the recent ups and downs, BTC’s price was high enough to continue driving increased competition among Bitcoin miners, who added capacity to the network in a bid to claim freshly minted coins for themselves. This has had the overall effect of pushing energy consumption ever upward, undermining the sense that there’s any seasonal variation.
Marc Bevand — an entrepreneur who has produced his own calculations on BTC energy consumption — largely agrees with this impression.
“We don’t notice seasonal variations because the network has been growing quite fast, so any — presumably small — seasonal variation is lost in the large amounts of hashrate capacity — and thus energy consumption — being added every month. For example, a year ago, the hashrate was at seven exahash/sec and has grown to 45 exahash/sec today.”
However, despite the overall impression that there has been one continuous increase in consumption, some subtle variations are observable in the data that de Vries has collected as part of the Bitcoin Energy Consumption Index.
For one, if you calculate the growth in consumption between the 2017 summer months and compare it to the three previous months, you’ll see a slackening in the overall rate of increase. For Feb. 10 to May 10 (Feb. 10, 2017 being the first date for which data is available), consumption increased by 33.1 percent:
- February 10 – 9.6TW/h
- May 10 – 12.8TW/h
But between June 1 and August 31 (meteorological summer), consumption increased by only 21.9 percent:
- June 1 –13.42TW/h
- August 31 – 16.37TW/h
What’s interesting about this is that Bitcoin’s price increased by 96 percent between June 1 and August 31, 2017:
- June 1, 2017 – $2,405
- August 31, 2017 – $4,714
By comparison, the price increase between the winter months of Feb. 10 and May 10 was ‘only’ 79 percent:
- February 10 – $978
- April 21 – $1,759
Put simply, BTC’s price grew faster over these three summer months of 2017, yet its energy cost grew more slowly.
Why? And what about the summer of 2018?
Well, in the three months between June 1 and Aug. 31, BTC energy consumption increased by only 5 percent:
- June 1 – 69.6TW/h
- August 31 – 73.1TW/h
Over the same period, BTC’s value sank by 6.3 percent. The thing is, its value dropped by a hefty 27 percent between March and May, during which time energy consumption actually increased by 31.6 percent. And between Dec. 1 and February 28, consumption increased by an impressive 69 percent, while the overall BTC value grew by only 7.8% between these three months (although they were big spikes over smaller time frames within this quarter).
As with the year before, 2018’s movements underline two things: a) that the growth in energy consumption slows down to an appreciable degree in the summer months, and that b) this slowdown can’t be correlated with price movements, particularly with regard to 2017’s figures. In 2017, energy consumption slowed down while price rises accelerated; in 2018, even though the price had sunk on Aug. 31 relative to its position on June 1, it was still 49 percent higher than it had been on Aug. 31, 2017. Such an annual difference should, in theory, provide a greater incentive for miners to mine Bitcoin and to increase their mining capacities, yet we nonetheless see that they eased up on their growth during the summer months.
Summer = higher electricity prices
The fact that BTC’s price doesn’t fully account for its energy consumption raises a conundrum. However, it’s one that’s solved via reference to the other biggest factor in Bitcoin’s use of electricity, which is the price of electricity itself. On the global level, electricity is generally more expensive in the summer, when there is greater demand for it, both from people turning on their air conditioners and from businesses — including mining farms — needing more energy for cooling.
For example, the United States Energy Information Administration — a branch of the U.S. Department of Energy — found in a 2013 review that energy consumption in the U.S. peaks in the summer for residential, commercial and industrial customers, with the variation ranging from 18 billion KW/h to 67 billion KW/h (compared to non-peak times). Similarly, in France and Germany, demand for energy during hot weather in June 2017 caused consumption to rise by 2GW and 4GW respectively. Meanwhile, China — home to some of the largest mining facilities in the world — has been facing the possibility of power shortages this summer, “as the nation’s distribution networks struggle to cope with soaring temperatures and the fastest power consumption growth in seven years.”
RMIT University’s Centre for Urban Research explained in a 2017 report on electricity pricing in Australia:
“During hot weather, the electricity sector aims to reduce peak electricity demand via ‘price signals’ — higher prices for electricity used at times when many households use air conditioning to cool their homes.”
As a single example, the U.S. Energy Information Administration notes in a January bulletin that wholesale electricity prices peaked at $55/MWh in California during August 2017, when they had been only $36/MWh in January of that year — amounting to a 52.7 percent increase.
It’s therefore clear that electricity demand and pricing tends to increase in summer, particularly on a global level — and particularly in China, where mining is most prevalent. By extension, this would explain why the increase in Bitcoin’s energy consumption also tends to level off slightly during the summer, since miners are reacting to increasing costs — and decreasing profitability — by temporarily reducing their capacity, at least in areas affected by hot summers.
This finding is backed up by the select few individuals who actually devote themselves to tracking Bitcoin’s energy consumption. Speaking to Cointelegraph, Ian Wright, the founder of Power Compare, confirmed that there isn’t really any significant or pronounced seasonality in Bitcoin’s energy consumption. However, what little seasonality exists is driven by the cost of electricity.
“If there is a seasonality effect, it would come down to electricity prices. So, for example, prices may come down in some areas with a lot of installed solar capacity when the sun is shining. Or it may go up in other areas that are hot, as more people turn on AC and increase demand.”
Marc Bevand, who doesn’t really see any significant variations in energy consumption, nonetheless also acknowledges that consumption levels are affected by profits.
“The energy consumption is driven mostly by increases of the price of Bitcoin. If miners make more profits, they will invest more capital in mining farms.”
While he doesn’t explicitly mention electricity here, this assessment is still consistent with the idea that seasonal electricity prices can affect consumption levels, since these prices will inevitably have an impact on profits.
This idea is also backed up by a May paper authored by CoinShares Research, in which Christopher Bendiksen and Samuel Gibbons investigated trends in the cost of mining Bitcoin. In particular, their research confirmed that mining companies are significantly influenced by seasonality:
“We also note that miner migration and/or price hikes occur during the dry season in China.”
Even though this paper didn’t describe any mining network reducing capacity, the fact that networks have a tendency to migrate whenever they can would suggest that, when they can’t migrate to an area with cheaper electricity, they may simply scale back. As the authors conclude:
“Some miners may have felt the squeeze during the market bottom, particularly if they were latecomers in terms of the modernity of their mining gear and/or operate in areas with comparatively higher electricity costs.”
Renewables
While what is above demonstrates that BTC energy consumption is lightly seasonal — in that the increase in capacity slips a little during the summer — there are two caveats worth addressing. The first, which is the less serious, is that the figures produced by Alex de Vries aren’t unanimously accepted by all those who track Bitcoin’s energy consumption. For instance, entrepreneur Marc Bevand constructed his own model for calculating BTC’s energy cost, finding that it was anything between 2.85TWh and 6.78TWh per year. This is considerably lower than de Vries’ first estimation of 9.6TW/h (for February 2017), which then grew to 32TWh for December, and then to 73TW/h for this August. It’s also lower than the estimation put forward by SetOcean co-founder Oscar LaFarga, who put the annual consumption at around 18.25TW/h.
Other commentators have put their estimations even higher than de Vries. However, even with this variation, de Vries recently noted that he used the BECI’s methodology to write a peer-reviewed paper — although it produced a lower estimate than that of BECI for overall production. He also notes that a Morgan Stanley report criticized Bevan’s approach, which allegedly underestimates the cost of mining networks for cooling, which alone can consume up to 30 percent or 40 percent of a network’s revenue. As such, this analysis has stuck with de Vries’ figures. What’s more, even if they are well into the upper range of possibility, the consistency of the methodology used for the BECI means that this would have little impact on the attempt to specifically follow increases and decreases in BTC’s energy consumption over time.
The second caveat, which is more significant, is that BTC’s modest seasonality may be weakened even further as the industry matures. Ian Wright says:
“[…] the price of Bitcoin relative to electricity prices is increasingly the main driver of consumption and is also driving a shift away from high-cost areas to places with lower prices.”
Marc Bevan describes a similar process:
“Miners also design their mining farms to run 24/7/365, so seasonal weather patterns don’t interrupt their mining operations.”
Here, Wright and Bevan are referring in part to the establishment of new mining centers in cooler nations such as Iceland, where Bitcoin mining is on course this year to burn more energy than all of the nation’s homes combined. Big mining companies, such as Bitmain, are increasingly flocking to areas with cheaper renewable energy and colder climates, such as Canada.
In the process, they’ll dilute the vague seasonality currently visible in energy consumption charts, enabling consumption to rise consistently for as long as Bitcoin’s price remains high and it retains its onerous proof-of-work (PoW) algorithm. And by doing this, mining firms will also help to reduce the impact Bitcoin is having on the environment. That said, an energy expert at the University of Pittsburgh recently observed that such firms are already making significant use of green energy sources, and that Bitcoin’s overall consumption is still negligible compared to that of the banking industry.
But until Bitcoin moves almost completely to renewables, its energy consumption will continue to exhibit some slight seasonality, easing its foot off the gas during the summer months just as the rest of world is doing the opposite. While this subtle decline might seem like a bad thing from the Bitcoin community’s perspective, it doesn’t appear to have any negative consequences in practice — except for maybe an increase in average confirmation time for transactions in the summer of 2017, something which hasn’t been a problem in 2018 due to the rolling out of the SegWit upgrade. In other words, Bitcoin’s capacity is growing very steadily, making it easier than ever before to send a transaction to its network and have it accepted.
