The Power Consumption of the Bitcoin Network Pt 1: Are We Destroying the Planet?

People are complaining how the Bitcoin network is based on a system of computers that consume an ever increasing amount of power to run the calculations needed to secure the network. They say it is a system that could only survive "in a space station with infinite energy, not a planet with finite resources". There have been many recent articles that try to analyze the consumption of the network, and make predictions on where it will go, but I believe they approach it from the wrong way and misunderstand the underlying causes of power consumption.

The two points that I see most are "The Bitcoin network will take more and more energy and accelerate global warming" and "Bitcoin will get more efficient as computers get more efficient". The reality is somewhere in the middle. It's not as bad as it seems, but its also not as good.

In this article I'll calculate the actual consumption of the Bitcoin network, and throw in some predictions of where it will go in the future. I'll also compare it to some other large energy consumers so we can see how it stacks up (or falls short!)

There are a few definitions needed before starting. First, when I say power ratio, I mean the amount of power it takes to get a set amount of hashes. The one I'll use most often is watts per gigahash (watts:gigahash). Second, I'm going to be calculating and jumping between orders of magnitude. Starting at kilo = one thousand, mega = one million, giga = one billion, tera = one trillion, peta = one quadrillion, exo = one quintillion. These get bigger by a factor of 1000 each time. These prefixes can apply as kilohash, kilowatt, gigahash, or gigawatts. A hash is a computer calculation that can be used for various applications, the Bitcoin network uses hashes to secure the blockchain. The speed of a Bitcoin mining machine is measured in hashes per second, or to be in the proper magnitude, gigahashes per second (billion hashes per second). The Difficulty of the network is a changing number that ensures the Bitcoins are released at a constant rate, even though the network is getting faster as more machines are added.

Seeing as Bitcoin is similar to a payment processor, it's only fair to compare it to other payment processors. Let's start with Ebay. I was trying to find PayPal, but couldn't get the information. Here's information on Ebay.

Very good! They're only at ~20 MW. Let's check some more internet giants. How about the biggest, Google? Here it is!

260 MW? That's pretty big! However, Google has many initiatives that are offsetting their global footprint to bring it down to 0.

Last but not least, let's check another giant, Facebook! Here we are.

"Facebook's server farms used 678 million kilowatt hours (kWh) of electricity in 2012," Quick note here, they measured that in kilowatt hours, which is a measure of power consumption. If you have a 100 watt light bulb, (0.1 KW) and run it for 1 hour, it has consumed 0.1 kilowatt hours. That's the comparison for kilowatts to kilowatt hours. So, we have to convert the Bitcoin network to kilowatt hours to make a comparison. The network runs for 24 hours a day, all year, so the Bitcoin network is using 110 MW * 24 hours * 365 days = 963 600 megawatt hours! Facebook is 678 000 000 kilowatt hours, convert that to megawatt hours is 678 000 megawatt hours in 2012. So Facebook is a bit smaller, about 70% of the bitcoin network.

*** Quick note, let's convert Facebook down to MW for easier comparisons. 678 000 MWh/(24*365) = 77.4 MW ***

Now, we have our scale:

If we try to guess the future power consumption by looking at the past, there was a stabilization at a difficulty of ~3 million. That was at a price of ~$20. Back then, peope just had graphics cards, which have a power consumption of about 650:1. Checking a calculator¹, using a difficulty of ~3 million, with 10 cents per kilowatt, and 1 gigahash consuming 650 watts, and a price of $20/BTC, we see that power consumption equaled about half of profits. Miners would make ~$3.20 per day, consuming ~$1.60 of power.

Extrapolating this into the future, we can assume people will continue to buy miners as long as they produce more Bitcoins in revenue than they consume in power. As the Bitcoin price goes higher, people will be willing to use less efficient machines. But let's assume a $1000 coin.

Everyday, 3600 Bitcoins are produced. One Block every 10 minutes, with 25 BTC per block. 3600 BTC per day, at $1000/ BTC, and using the "1/2 profits to electricity" metric, miners would then make 3.6 million dollars a day, and spend 1.8 million in power. 1.8 million dollars, at the average price of $0.10 per kilowatt, is 18 million kilowatt hours, or 18 000 megawatt hours per day. Divide by 24 hours in a day, to get 750 MW. This would then shoot us past Google for power consumption, which is at 260 MW. However, in the future, the block rewards will half every 4 years. In 2017, it will half to 12.5 BTC, then 6.25 BTC in 2021... Theoretically the rising BTC price will keep people wanting to mine. But this will affect how much money the network generates per day.

If and when Bitcoin gets to this coverage, this much network power, Bitcoin will be a secure player in the world. It will be covering lots of payments daily, and will be a huge force acting on the world. At this scale, there will be even more companies making industrial scale mining rigs. At this point, it would be fair to compare power consumption to gold mining companies, as well as the federal reserve, with them printing money. Here's data for gold. It takes about 25 kWh to produce 1 gram of gold. Times that by 28.34 grams per ounce to get 708 kWh per oz of gold. Compared to Bitcoin right now, it takes 300 kWh per Bitcoin. (45 megawatts = 45 000 kilowatts * 24 hours a day / 3 600 BTC per day).

In the future, there is an estimated 18 000 000 kilowatt hours to generate 3600 BTC, which gives us a massive 5000 kilowatt hours per BTC. Once the network reaches this scale, it may run into some issues. But again, once it hits this scale, it will be competing with banks, the federal reserve, gold miners... We need more data to make more accurate comparisons.

This article has a great comparison of the bank offices of the major banks.

The conclusion is that the top 20 banks offices consume 2628 MW! That's almost 5 times the current Bitcoin network.

This number can also be used to predict difficulty. 750 MW of power, using the 5:1 watts:gigahash ratio, gives a total network of 150 petahashes. 150 petahashes would give a total difficulty of 21.43 billion. We can run these numbers through a calculator. Since we're in the future, we can assume use of the best miners. Right now, the KNC Neptune provides 3 000 gigahashes for a mere 2100 watts. Checking the calculator, 3 000 gigahashes at 21.43 billion difficulty, with a $1000 coin, gives you about $65 a day. The Neptune costs $14000, so would someone make the investment of $14 000 to make $65 a day? That gives you a return on investment of 215 days. It could happen. But nowadays, people are looking for ROI in 3 months or less. That is because the difficulty is going nuts right now, in the future if it stabilizes they may be willing to go for longer returns.

As of early 2017

The network is sitting around 2.7 exohashes/sec (2,705,698,350 GH/s) and there was a block halving which reduced the bitcoin reward from 25 BTC/block to 12.5 BTC/block. Luckily, the mining efficiency has increased as well!

The previous estimates put the average power consumption ratio (watts:gigahashes) at 5:1. Currently, it's closer to 0.2:1. The best miners can get 0.1:1, but we'll sit a bit above to allow for error. For a network of 2.7 exohashes, that gives us a power consumption of (2.7 exohashes = 2 700 000 terahashes * .2 kilowatts = 540 000 kW) 540 MW. That gives us a new number to compare with. I've tried to find new numbers for the other internet companies I compared to as well.

Google doesn't disclose data on it's power consumption, but this article can give us a good idea. It claims 4,402,836 MWh in 2014, which converts to 502 MW. It also claimed a 40% reduction with it's DeepMind AI, so we can add that in and assume Google has 40% reductions in future years (301 MW). These new numbers have been added to the comparison chart above.

Since we're getting to bigger numbers, we need bigger comparisons! "US data centers consumed about 70 billion kilowatt-hours of electricity in 2014"² That is 70 000 000 000 kWh, which converts to 8000 MW. In that article, they extrapolate data and estimate that even though the demand for data centers grow and more computing is needed, the improvements in energy efficiency means that the total power consumed will stay roughly the same.

We can check this new data with the current bitcoin price. It's hovering around $900 USD, so let's see if the "half profits" metric from before fits. At 540 MW, and a US average price of $0.1/kWh people are spending 1.3 million (540 MW = 540 000 kW *24 = 12 960 000 kWh * $0.1 = 1.29 million) dollars a day on the Bitcoin network. This is a bit smaller than the previous prediction (~1.8 million dollars, 750 MW). However, the main difference is the block halving, and the price change. Miners are only producing 12.5 BTC/block, which is 1800 BTC per day, equaling 1.62 million dollars a day. Miners are spending 1.3 million to make 1.6 million, way over the "half profits" metric that we thought would hold. However, this does also assume a 0.2:1 watts:gigahash ratio, which we knew was an overshot error. If we take the best miners on the market (0.1:1, Antminer S9) miners are only spending $650 000, which easily puts us under the "half profits" metric.

However, this is all assuming US averages. Bitcoin is a worldwide phenomenon, and electricity costs different amounts in other places. Also there are places where people set up miners with free electricity *cough*China*cough* which would mess with the data.

All things considered, Bitcoin network consumes less electricity than predicted, and still far less than an "environmentally destroying" amount.

As of late 2017

The price has now gone insane, which made lots of old miners turn on their old machines, which would make the network take more energy. Let's break down the numbers.

Current price is ~$16 000 USD, and rising. Luckily, the price moves faster than the hardware can keep up, so just because the price spikes doesn't mean the electricity will spike as well. Even still, the electricity consumption will grow. So, a $16 000 coin means there is just under 29 million dollars of Bitcoin being produced daily. Half profits means people would spend ~15 million on electricity, which works out to 6226 MW! This spikes up like crazy, boosting us way beyond all the predictions. However, power is still below all the US data centers, though it's getting close...

But, that's the prediction from the electricity side. Let's look at it from the miners side. The network is sitting at 8.66 exohashes/sec (8,661,633,671 GH/s). The best miners are able to run at 0.1 Watts/gigahash, but since the price is so high we can assume some old miners have been turned back on, and the average is at 0.2 watts/gigahash. Running those numbers, there is a network size of 1732 MW (8.66 exohashes = 8 660 000 terahashes * .2 kilowatts = 1 732 000 kW = 1732 MW).

That's a significant jump, but not too massive. Here are some more comparisons. The top 10 power plants produce between 6 400 MW and 22 500 MW. They are all hydro plants, except #6 which is nuclear.

On a greener energy side, Tesla just installed a 100MW battery in Australia³. We'd need a few of those to power the Bitcoin network...

Using global energy data⁴, we can calculate that the Bitcoin network currently uses ~0.0098% of the worlds energy. Or, it consumes 0.039% of the total US energy consumption. I wouldn't be too worried about it until it approaches 0.1 or 1% of global energy usage, which would require a 100x increase, putting the BTC value at close to 2 million per coin.

In conclusion, the Bitcoin network may consume a lot of power, but it doesn't take a massive, environment destroying amount. We'll have to see what the future holds. The offset of people building more and more power efficient miners is that people are willing to buy more and more of them to make money. Who know's what will happen? Part 2 of this article will explain how even though Bitcoin and other cryptocurrency networks are consuming lots of power, they can use excess generated power, which would lead to less global waste and make the planets energy consumption more efficient.

Originally published on Medium. More writing