One of the criticisms of blockchain technology center on power consumption and the cost of mining. The electricity consumed in mining operations is high and continues to grow as mining difficulty increases in response to increasing processing power. Some think that blockchains is wasteful technology that is not compliant with a 21st century outlook for a green economy. Others accept it as the price to achieve consensus and trust in decentralized networks. Following recent work in this space, I like to share a perspective by making a comparison to the energy used in communication systems.
Bitcoin mining vs. mobile base stations
My estimates for bitcoin mining power consumption is around 7,000 MW, a average of a floor of 1,200 MW and ceiling of 12,000 MW. Many external sources point to values in the same range (at the time of writing, this example places power consumption on par with that used in Singapore). By comparison, I estimate that all the mobile base stations use around 12,100 MW, of which around 20% is in China. The power consumption for bitcoin mining has been increasing at a relatively fast pace. On the other hand, power comparison of mobile base station increases at around 2% annual. Based on this, the power used for bitcoin mining is set to overtake that used by mobile base stations this year, by late fourth quarter.
Power consumption in data centers
Data centers power the Internet, and in 2011, Google used around 260 MW to power its data centers. Today, around 8,000 MW power US data centers alone. Data center power consumption is so important that Google acquired DeepMind, an AI company, for over $500 m to optimize and reduce energy utilization in its data centers.
Power generation benchmarks
To put all this into perspective, the average hydro-electric power plant has 1,000 MW capacity with the largest at the Three Gorges Dam in China peaking at 22,500 MW. In contrast, the average nuclear power plant generates around 4,000 MW with the largest Japan’s Kashiwazaki-Kariwa peaking at 8,000 MW. Coal plants generate on average 2,000 MW with the largest in Taichung, Taiwan peaking at 5,700 MW. It takes about 9 average hydro electric plants to run the entire install base of wireless base stations, and soon, a similar number to run all of bitcoin mining operations.
Bitcoin mining power consumption in perspective.
Value is in the eye of the beholder!
Whether this level of power consumption is too high can only be measured with the economic utility of the application. There are no vocal, public concerns on the energy used in communication systems. Most people are oblivious to the details and readily accept the price of the service. However, the industry recognizes that power consumption could be improved, especially as networks grow ever larger. Specifically, there’s much wasted power due to transceivers operating continuously irrespective of the data being transmitted. This is an area left to 5G technology to address. In the meantime, Internet players keep power consumption used in data centers secret. But as the Google acquisition of DeepMind demonstrate, it is an item high on the priority list.
New solutions for mining challenge
Blockchain mining is another story all together. Today, most mining happen in parts of China, Mongolia and other countries where power is cheap and climate is cool. The crackdown on bitcoin in China left miners eager to explore options to move elsewhere. Russia is particularly active in luring some of these miners.
Blockchains remains a nascent technology yet to break out beyond the crypto currency application best exemplified by bitcoin. The utility of blockchain is yet to be validated beyond the hype that bitcoin generated. The proof-of-work based on the SHA256 cryptographic hash function is specific to the bitcoin blockchain, but this is not the sole implementation. In fact, Ethereum plans to switch to a proof-of-stake technique that will reduce power consumption.
Energy-efficient mining puzzles remain one of the hottest areas of research in blockchains and there are many different proposal. Therefore, while blockchains consume a growing amount of power, we cannot extrapolate too long into the future as the technology is evolving and will become more efficient as different techniques are introduced.
Thought-provoking side note
A side note to this article – for those interested in the future of bitcoin. The energy consumption is a prism to look at the future of bitcoin. Consider that the level of difficulty and power consumption rises at higher pace than it is possible to enhance the power efficiency of mining operations. Then factor into that diminishing mining rewards over time which leads to higher transaction fees to compensate miners for revenue shortfall. The value of bitcoin should appreciate to cover any revenue shortfall for miners. Should miners not find mining operation profitable, they will exit which will threaten the viability of the systems and could potentially lead to its collapse!
>they will exit which will eventually lead to the system collapsing! But this is a thought for another post.
I am not an expert, but isn’t that the point of the cypto currency algorithm. That as the mining operations become cost prohibitive (in relation to the rewards), there will reach a plateau in the growth of the coins, which means that we will deal with a money supply that is stable. That means that unlike currency, these things can’t be printed (mined) by governments. So, that means inflation is controlled automatically. hence, (crypto)currency stability as opposed to today’s print-your-way-out-of-trouble economy models. Interested in your thoughts …
You’re correct. In the case that miners don’t make enough in rewards or transaction fees to cover their expenses, they will leave. This can happen if demand on the crypto currency is low such that it does not support a valuation high enough to justify mining expenses which keep on increasing. At some point as miners exit, the difficulty will reduce correspondingly potentially making the reward and fees worth the effort. In theory, the system should keep adapting as such, but I am not sure if in practice it will work this way! Will be interesting to see if anyone modelled such scenarios.