The Internet’s energy consumption

The internet has dramatically changed the way we access information. Also the way we use the internet is evolving quickly which results in increasing amounts of data being exchanged. At the beginning messengers were text-based tools and we used Google like an online encyclopedia. Nowadays we stream videos in HD (soon 4k, later 8k) on demand which requires point-to-point connections rather than traditional broadcasting, we search DIY videos to learn new things, share our photos on Instagram and keep in touch with our friends on Facebook.
All this requires data to be stored, processed and transmitted which consumes energy. In the following some considerations on the energy consumption of the internet.

It’s the physics, Stupid!

There are physical limitations regarding the minimum energy per transmitted or processed bit which is measured in Joule per bit (J/bit):

  • Computation: Landauer’s principle: At  room temperature (20°C), the Landauer limit represents an energy of approximately 2,75zJ/bit which equals to 2,75 x 10-21J/bit. Theoretically, room‑temperature computer memory operating at the Landauer limit could be changed at a rate of one billion bits per second with energy being converted to heat in the memory media at the rate of only 2,85 trillionths of a watt (that is, at a rate of only 2,85pJ/s). Modern computers use millions of times as much energy per second.[1]
  • Wireless data transmission: Shannon’s limits puts the physical limit on the wireless data transmission to around 1fJ/bit which equals to 10-15J/bit.: For practical number of antennas and channel gains, we can rather hope to reach the order of under 1pJ/bit which equals to 10-12J/bit in future systems.[2]
  • Wired data transmission: In principle underlies the Landauer’s principle, in practice three orders of magnitude more energy efficient than wireless, around 1fJ/bit.

Please note that both wireless and wired data transmission limits only consider the actual transport of data, totally ignoring any processing needed to encoding and decoding the data payload with different protocols and rooting it through many nodes from source to sink. With current technology an end-to-end internet data transmission of data is in the ballpark of 25 – 75µJ/bit[3]

Google, Facebook and Co.

Here some examples of famous internet services. One might wonder, why the presented energy consumption consider only to the delivery of data, but not the energy consumption of the end-user devices. This is because typically the end-devices can be neglected. For example with YouTube videos, delivering the content consumes 200mWh per minute of streaming video, while playing the video on your mobile consumes one order of magnitude less, or 25mWh per minute of playing video.[4]

Alphabet (Google’s parent company)

The energy needed to power a Google Search was 0,0003kWh in 2017.[5] In 2018 Google received over 63 000 searches per second on any given day. That’s the average figure of how many people used Google a day, which translates into at least 2 trillion searches per year, 3,8 million searches per minute, 228 million searches per hour, and 5,6 billion searches per day.[6] If we consider the energy per Google search we can calculate the energy consumption of Google searches for 2018 as follows: 0,0003kWh x 5,6 109 x 365 = 0,613TWh. But it is interesting that the total energy consumption of Alphabet was 22,776TWh in 2017, so it seems that Google searches consume only a tiny fraction of the total used energy of 2,7%. What are they doing with all that extra energy?!?

Total consumption of Alphabet:

2010: 2,26TWh[7], 2015: 5,70TWh[8], 2017: 22,78TWh[9]

Facebook

According to Facebook statistics there were 2,32 billion monthly active users (MAU) as of December 31, 2018[10].

Total consumption of Facebook:

2011: 0,532TWh, 2012: 0,704TWh, 2013: 0,822TWh, 2014: 1,035TWh, 2015: 1,31TWh, 2016: 1,83TWh, 2017: 2,46TWh[10]

YouTube

YouTube reported the following staggering statistics in 2019:[11]

  • Over 1,9 billion logged-in users visit YouTube each month, and every day people watch over a billion hours of video and generate billions of views.
  • More than 70% of YouTube watch time comes from mobile devices.
  • The average mobile viewing session lasts more than 40 minutes.
  • YouTube is the world’s second largest search engine and third most visited site after Google and Facebook
  • 400 hours of video are uploaded to YouTube every minute

Together with the information that Streaming one minute of YouTube, eats up 0,0002kWh[12] one can calculate the energy consumption for 2018 as follows: 0,0002kWh x 109 x 365 = 4,38TWh

Bitcoins

There are also more hidden online services that have a massive need for energy, like cryptocurrencies. The energy used for mining is highly volatile and hard to predict and in the case of Bitcoin it is significant.

Total consumption of Bitcoin:

2017: 32.4TWh[13], 2018: 62.3TWh[14], 2019: 50TWh[15]

The big picture[16]

Let’s have a look what the above means in the big picture, related to the global electricity consumption. Also we need to consider the end-to-end energy need of our Information and Communication Technology (ICT) which consists of consumer devices (including their production) for end-user interaction, wired and wireless network to transport data to and from the end-users and data centers (here Google, Facebook and Co are included) that store/process/provide the data.

In 2018, the global consumption was 23 809TWh out of which 2 476TWh or 10% were ICT, the trend suggests that in 2030, the global consumption will be around 40 000TWh out of which around 8 000TWh or 20% were ICT. Remarkably the predictions show that the energy need of ICT grows much faster (exponential) than the global energy consumption (linear) and will account for 20% in 2030. This also means that ICT has an big impact on the global increase in consumption. It is also surprising that within ICT a significant consumption growth of data centers and wired networks is expected, while mobile data for example is predicted to stagnate despite the 5G plans of connecting billions of devices to the internet.

<elvis, 3.4.2019>

Leave a Reply

Your email address will not be published. Required fields are marked *