Digitalisation: A drain on resources?

Symbolic picture for the article. The link opens the image in a large view.

Is digitalisation really more sustainable?

In the early years of digitalisation, hardly anyone could imagine the impact information technology would have on our daily lives. In the 1940’s, IBM chairman Thomas Watson predicted, ‘I think there is a world market for maybe five computers.’ Nowadays, if you add together the computers, tablets and smartphones in any typical two-person household in Germany, the chances are that you will already come to more than five computers. And that’s before you even start to consider devices such as smart watches, smart TVs or robot vacuums.

However, there is also a downside to our digital aids. Electricity consumption for digital applications is steadily on the rise. A single Google search is estimated to cost roughly 0.3 watt-hours of electricity. On a global scale, Google alone receives approximately 2.4 million search queries per second, consuming approximately 720 kilowatt hours (kWh) of energy. That is roughly one third of the annual electricity consumption of a single person household in Germany.

Annual greenhouse gas emissions caused by digitalisation in Germany are considerable, as shown in the KfW report ‘Deutschland auf dem Weg zur Klimaneutralität: Welche Chancen und Risiken ergeben sich durch die Digitalisierung‘ (Germany on the road to climate neutrality: Opportunities and risks of digitalisation). The report states that digital applications accounted for approximately 34 million tonnes of CO2 emissions in 2020. This also includes emissions caused by the production of end devices and the digital infrastructure. The figures show that digital applications account for a share of just over four percent of Germany’s total emissions of greenhouse gases. Based on the average private use of digital devices, the digital footprint per person and year amounts to approximately 740 kilograms of CO2. If devices are used intensively, the amount rises to approximately 1050 kilograms of CO2 per person per year.

Saving energy when encoding videos

Einzelne Bilder kommen aus dem Hintergrund nach vorne geflogen.

The rapid growth in internet use and the resulting energy consumption has been driven predominantly by the move towards watching videos online. Even before the coronavirus pandemic it accounted for a total of 80 percent of traffic. Most users access streaming platforms, followed by pornography, video platforms and videos on social media channels. ‘Growing data streams and the connected emissions are a global problem,’ comments Dr. Christian Herglotz from the Chair of Multimedia Communications and Signal Processing at FAU.

Converting electricity consumption by end devices for streaming videos into CO2 equivalents, in other words calculating the resulting CO2 emissions, is a complex matter. Depending on the source of energy, CO2 emissions per kilowatt hour vary from approximately 13 grams for hydroelectric power to up to 1230 grams for coal. The decisive variable is the availability of renewable energy, but that depends on the weather. Another variable influencing energy consumption for streaming from data centres, including the final leg of transmission by Wi-Fi, is the end device itself, whether it is a smartphone, a tablet, a laptop, a 40 or 50 inch TV or a PC.

The energy consumption of an end device for streaming also has several dimensions,  as Herglotz and his team demonstrated in their laboratory using a smartphone. They recorded the power needed to decode and play the video, the brightness of the screen and audio requirements. Their measurements show that when watching an HD video, most power is consumed for idle time, for video streaming, and for the screen.

From a technical point of view, video codecs, special algorithms for coding and decoding, allow providers to compress their video data for transport over the internet and end devices to decode the compressed data they receive. The necessary hardware chips in the end devices are already highly efficient and significantly reduce the energy required. However, if an older end device does not have the suitable microchip, then it is the processor in the end device that is responsible for decryption. This may require twenty or even one hundred times as much energy.

Herglotz is now working on an idea for making this transmission process more energy-efficient. He believes including this additional feature for regulating energy requirements in the codecs is a pioneering approach. In future, a mobile phone could automatically intervene in the transmission process and request a stream that uses less energy whilst still providing a film of the same quality. This is the first time that the dimension of decoder energy has been taken into consideration alongside the two existing principles of video compression: minimum bit rate and maximum video quality. In practice, the smartphone would inform the data centre that the transfer of the remaining 30 minutes of the video should be adjusted in light of the remaining battery life of 20 minutes.

Compared to countries

The Cambridge Bitcoin Electricity Consumption Index estimates that the cryptocurrency Bitcoin consumes 144 tera-watt hours (TWh) per year, that would be more than the energy consumed by the whole of the Netherlands. Others assess the energy requirements to be less than 118 TWh, equivalent to the energy consumed by Peru.

Crypto currency: Data protection or energy efficiency?

Technical progress means that electricity consumption per transferred gigabyte actually halves every two years. This long-term trend towards more energy efficiency is counteracted, however, by what is known as the rebound effect. More people streaming videos for a longer period of time require greater amounts of energy and therefore emit more CO2 as a result. The same applies to audio data, images or the video conferences required as a consequence of the coronavirus pandemic.

Another digital energy drainer was brought to the attention of the general public this year by Tesla boss Elon Musk: the cryptocurrency Bitcoin. First of all, he announced to his nearly 60 million followers on Twitter that the electric car manufacturer intended to accept Bitcoins as payment. However, he then back-pedalled: Due to the energy consumption entailed by the Bitcoin blockchain, Musk stated that he would only use the currency once sustainable energy was available.

The oldest and most well-known cryptocurrency guzzles energy due to its special consensus algorithm. Part of this procedure is based on ‘proof of work’ (PoW). This is used to create new blocks and consumes approximately 90 percent of the energy.

Prof. Dr. Dominique Schröder from the Chair of Applied Cryptology at FAU is also keeping an eye on the high energy consumption. However, he also believes that there are advantages, particularly the fact that the Bitcoin network has worked without interruption from the outset thanks to its decentralised structure. ‘That is certainly not the norm in the IT world.’ Further advantages he sees are the security aspect, the prevention of political intervention and finally also the high level of transparency of the transactions.

In addition, criticism of the virtual Bitcoin currency fails to take into account the costs incurred by physical money, stresses Schröder. Worldwide, it is estimated that the production costs for paper currency are as high as five terawatts every year. The global energy requirements of the banking sector account for an additional 100 terawatts per year.


Before the coronavirus pandemic, the French think tank The Shift Project calculated that digital technologies were responsible for four percent of global greenhouse emissions. In 2025, emissions from digital technologies could already account for eight percent of the total. The production of computers, televisions and other end devices accounts for almost half of greenhouse emissions. Global civilian aviation generates almost three percent.

When it comes to the CO2 footprint of cryptocurrencies, other digital currencies such as Cardano have a much better track record. This mining algorithm is not based on delivered computing performance, but rather on the stake the miner has in the currency, known as the ‘proof of stake’. In addition, ‘off-chain transactions’ are becoming more and more popular for cryptocurrencies. They are no longer logged directly in a crypto blockchain (‘on-chain’) but rather run via a bypass or extension. This accelerates the payment process using virtual money at the same time as reducing computing time and consequently also energy consumption. ‘This off-chain payment helps to keep the huge energy consumption in check,’ says Schröder. Off-chain payment therefore makes it easier to pay smaller amounts as well. That was the requirement that had to be met before El Salvador recognised Bitcoin as a legal means of payment in 2021.

Schröder, who focuses predominantly on the issue of individual privacy, believes that blockchains or cryptocurrencies reflect an essential conflict: the conflict between the issues of increased privacy in the digital space and increased efficiency. Efficiency in relation to cryptocurrencies means conducting fewer operations and thereby saving energy. ‘Unfortunately, in practice efficiency tends to win over individual data security,’ comments Schröder. This rivalry accompanies digitalisation in all its forms, whether in private households, companies or society as a whole.

Energieeffiziente KI

Bitcoin-Münze mit Spitzhacke zwischen Goldbrocken.

Blockchain technology is also becoming established as the method of choice in industrial models. Securely saved data in an unalterable form can be used, for example, to track the progress of individual raw materials along the chain, from transportation routes, to production, to when they are sold as a finished product. In the diamond industry, for example, this could prevent ‘blood diamonds’ from finding their way into the retail trade, and in the pharmaceuticals branch it could help curb the lucrative trade with counterfeit products.

Industrial blockchain products do without the PoW principle, which makes it less energy-intensive to create new blocks. Instead the companies take a decision concerning the next block in their block chain according to hierarchical rules, using Quorum. There are no valid figures about the energy required for this more energy-efficient procedure, but Schröder estimates that it saves approximately 90 percent compared to Bitcoin, as much less computing power is required.

Similar energy efficiency is also possible with an artificial intelligence (AI) solution from Prof. Dr. Dietmar Fey, Chair of Computer Architecture at FAU. He develops microchips that use solid-state memory, known as RRAMs. As these programmable circuits are capable of remembering their settings even when not connected to power, they can continue to work immediately after being switched back on. This means that the RRAMs can record data whilst the circuit for the AI algorithm is in an energy-saving hibernation mode. If the circuit is wakened, the required weights are immediately available for the neural network. Compared to systems that are constantly on standby, this can lead to energy savings of up to 95 percent.

Fey uses ternary values to save the weights efficiently. Ternary values use the digits zero, one and minus one and allow learning processes to take place, similarly to in a neural network. Their decisive advantage is that they require less memory, which then in turn has a positive knock-on effect on energy efficiency.

Negative side-effects

More negative aspects of the Bitcoin hype: Computers used for mining generally only have a life-span of two years. Not only that, Bitcoin miners use up roughly one quarter of all semiconductors produced in a year. That acerbates the global shortage of chips required to manufacture cars, smartphones or games consoles.

It all adds up

This approach was simulated in a joint project between the Chair of Computer Architecture and the Institute of Electronics Engineering led by Prof. Dr. Robert Weigl with the assistance of Marc Reichenbach and Amelie Hagelauer, and together with the Fraunhofer IIS in Erlangen and Dresden. The consortium was awarded first prize in the pilot innovation competition ‘Energy-efficient AI systems’ from the Federal Ministry of Education and Research. The model recognises cardiac arrhythmias and atrial fibrillation with an accuracy of at least 90 percent from ongoing ECG data and uses very little energy in the process.

Großer Serverraum.

The prize money is to be used for a subsequent project to provide new software tools that will improve AI training to allow the implementation of efficient time series analysis of continuous data streams in other applications. At the same time, energy minimisation is taken into account. Fey sees a lot of potential for different applications both in medicine and in industrial automation or quality management.

Fey takes a more sober view of the impact it will have on saving energy: ‘It all adds up.’ In absolute terms, only small energy savings will be made. However, if sensors using such technology are used on a major scale, relevant savings could be made.

Innovative approaches such as these help resolve the dilemma of more digitalisation or more sustainability. Slowing digitalisation is not an option for achieving our goal of a planet with lower CO2 emissions. Instead, we need to find a way to uncouple digital applications from increasing greenhouse gas emissions.

About the author

Thomas Tjiang has worked as a freelance journalist since completing his master’s degree at FAU. His main topics are business, economics and social issues.

FAU research magazine friedrich

Auf dem Cover des FAU-Magazins sind Spielzeug-Windräder in gelb, rosa, grün und blau zu sehen

This article first appeared in our research magazine friedrich. You can order the print issue (only available in German) free of charge at

All articles