Reading time: 11 minutes
Translation by AB – April 17, 2020
Science, any science, is without conscience or limits.
André Comte-Sponville – French philosopher
Digital and Energy
The digital world has been recognized for a good ten years as an energy consumer with a sensitive ecological footprint. In these times of disruption of the earth system linked ultimately to our energy consumption, this can become at least problematic for the image of a sector that promises to solve our problems of employment, health, overcrowding and… ecology!
Digital growth must therefore be sustainable: economy and politics must maintain the delusion of a “green” digital world. Everything also shows that from, a personal point of view, we do not feel that digital is a problem; on the contrary, it seems to us “green”, “smart” and a contributor to our eco-friendly attitude (bicycle, electric scooter, home automation, sorting of waste, fight against junk food, minimization of trips, etc.).
However, it is incredibly difficult to obtain consistent and comprehensive data on the actual power consumption of the digital world, for several reasons. First, as we have just seen, the image of the sector must be preserved and their very controlled communication hesitates between transparency and, let’s say, restraint. Then, this consumption is global and diffuse and the sources of information more or less available and homogeneous according to the countries. Finally, most of the consumption in the digital world, either is not due to end users or is not carried out at their local level: a Google request consumes energy distributed between the computer, the WIFI box, routers, Google servers, etc. but also succeeds because all these devices, permanently on, consume to be simply available1.
Electric consumption is therefore difficult to estimate because globally spread out in billions of “always on” subsystems and activated by billions of agents, human and artificial. Studies, essentially American to our knowledge, have been carried out and are still sufficient to assess a few orders of magnitude and the possible evolution of the situation in the medium term.
2007 – Digital saved from its polluter fate?
The digital world, the internet in particular, experienced inflationary growth during the 1995-2005 decade (the “internet bubble” was a financial bubble that changed almost nothing in terms of physical growth and therefore energy growth in the digital world). It was only in 2007 that the first report was published about a possible problem concerning the energy consumption of what specialists also call “ICT” (“Information and Communications Technology” domain). The findings of this report by the Gartner Group were presented as follows2 :
The ICT industry accounts for about 2% of total carbon dioxide (CO2) emissions, about as much as aviation, according to new estimates from Gartner, Inc. Despite the overall environmental value of IT , Gartner believes that this is not sustainable.
Let us dwell a little on this formulation. First, Gartner causes a feeling by choosing to evoke, not the invisible electrical consumption, but rather the almost observable releases of greenhouse gases. The comparison with aviation is not chosen at random (and will be repeated for many years by all the media): who does not see, does not hear, does not feel that an airplane is polluting dreadfully? Second, the “overall environmental value of IT” is undeniable, isn’t it? Finally, concluding that the trend is not “sustainable” sounds more like a (political) trigger to a strategic industry than a warning from Greenpeace. In “Gartnerian” language, that means: focus on this! There is a great business to do on the subject of sustainability in the digital world.
Indeed, the report opens a new period of expansion of ICT allowed by the banner “green IT”. Like AI players who secrete their own ethics, IT repaints itself in green, fortunately makes undeniable progress in terms of consumption, in particular carbon-free energy, and then allows itself to consume without counting and above all without being challenged by civil society. The period 2005-2015 was green and smart.
2008 – The climate saved by Digital?
This remarkable strategic movement is relayed by another actor, the GeSI (acronym for the evocative “Global e-Sustainability Initiative“), a group of companies in the sector and organizations founded in the early 2000s and whose mission is thus described3 :
The GeSI is a leading source of impartial information, resources and best practices for achieving integrated social and environmental sustainability through ICT.
Just after the Gartner’s report in 2007, GeSI published in 2008 its own analysis, a sort of roadmap for the industry, entitled “SMART2020: Enabling the low carbon economy in the information age“4. The main idea is as follows: the carbon footprint of ICT will more than double by 2020 to reach 1.4 GtCO2e5 (i.e. 3% to 4% of global discharges, which are around 40 GtCO2e), but ICT will contribute to an overall reduction in emissions of 7.8 GtCO2e, a positive balance of 6.4 GtCO2e (Gartner’s so-called “overall environmental value of IT“). Thus, the ICT roadmap is clear: to become a net contributor to the reduction of greenhouse gas emissions. An essential principle seems to be commonly accepted today6 :
Data collected by scientists from the Global Carbon Project [GCP] suggest […] that decoupling between economic growth and growth in carbon emissions is possible.
The GeSI goes even further: the economic growth of ICT – and therefore of its final consumption – is a condition for reducing overall carbon emissions. How? Essentially by “smartifying” (making intelligent) all our artefacts and infrastructure networks (cities, roads, electrical networks…) and allowing the optimization of their uses with, as a consequence, a reduction of emissions linked to transport, buildings (heating / air conditioning), industry, agriculture, etc.
We have therefore had a great decade of “green” and virtuous growth in the digital world, motivated by the reports of Gartner and GeSI, which will be widely reused, in particular by the European Commission or COP21.
2015 – The ecological triumph of digital
In 2015, the year of COP21 in Paris, the GeSI updates its report and two engineers, Galenbe and Caseau, publish a study in which they analyze the energy consumption in the digital world7.
Here is an order of magnitude to begin with: the world consumption of electricity is around 20,000 TWh8 (500 TWh for France), or 15% to 20% of world energy consumption (120,000 TWh).
The physical substrate of the digital world consists of three sets: terminal equipment (smartphones, personal computers, connected objects – sensors, cameras …), communication networks (WIFI equipment, cables, routers …) and data centers (“computing” servers). The total electrical consumption of each of these assemblies is in principle distributed over the entire life cycle, from manufacturing to destruction, including use of course.Galenbe and Caseau estimate that ICT consumed around 4.7% of the world’s electricity in 2012, just over 900 TWh. The main consumers are terminals (55% of total consumption) followed by networks (25%) and, finally, data centers (20%). If the data centers impress by their gigantism, they are also those on which the greatest progress is made in terms of consumption and carbon-free energy 9. The big players control their consumption (they have the means to invest to reduce it) and technical progress is such that in 2015 all the authors confirm the good newsThe big players control their consumption (they have the means to invest to reduce it) and technical progress is such that in 2015 all the authors confirm the good news:
We see no evidence that the ICT footprint is likely to grow unsustainably large.
Nearly ten years after the Gartner report, not only does ICT’s electricity consumption seem objectively manageable and “sustainable”, but the digital world has become necessary for the climate rescue10.
2018 – The impossible consumption forecast
In the long term, nothing is less certain. We would rather say: “we see no evidence that the ICT footprint is not likely to grow unsustainably large”. No one is now offering credible arguments for long-term ecological benefits in the digital world. There are not even any possible arguments today in 2018 and there are at least three reasons for this.
First, a widely accepted argument will be put to the test. Here it is, as expressed by the French association négaWatt:
The time available for engaging in digital activities is not infinitely extendable. Saturation and substitution phenomena are starting to be noticeable, and the consumption of the devices no longer adds up mathematically. For example, sales of desktop computers have been at “half-mast” for several years, because we replace them with laptops (less energy-consuming) and we use them less because smartphones (even less energy-consuming) take over.
However, the Technological System does not tend to work this way: to human consumption, certainly not “infinitely expandable” is now added the consumption of more and more intelligent and autonomous artefacts, consumption in principle without limits. In short, artificial intelligence is reshuffling the cards: to human society is added a “society of machines” which claims its energy share and which, if one can personalize in this way, is not “concerned” by reducing the ecological footprint. Gone are the days when it was possible to make consumption forecasts by roughly multiplying the quantity of artefacts (cars, refrigerators, planes, internet boxes, etc.) by the world’s population. The digital world decouples these two variables which become foreign to each other.
Second, the “smartification” movement is crisscrossing the physical world with billions of connected objects (smart city, smart grids, etc.). Studies have been carried out about the additional electrical consumption generated by these objects (so, there is a matter!). Remarkable optimizations are imagined, but this gigantic artificial mesh is far from having accomplished the path towards any naturalness, which would consist, for example, of effective feedback with the natural environment. The consumption of these objects will always be additional, increased tenfold by an energy-consuming communication infrastructure (waves), multiplied by a gray energy of incessant remanufacturing (obsolescence), not to mention the drain on natural resources (metals, rare earths, etc.).
Third, forecasts (Gartner, GeSI, etc.) are made simply by extrapolating trends, all other things being equal. The history of the digital world is still in its infancy. It helps to change our lifestyles, the nature and the quantity of our needs. Consequently, as much as it is possible to “foresee” the impact of the ICT on the existing (transport, consumption of material goods, etc.) and, of course, to consider it as positive, the more impossible to foresee the new needs and uses generated by the digital world itself (no offense to some prophets …). Who could for example foresee the explosion of cryptocurrencies, bitcoin for example, a solution which has no equivalent in the physical world from which we could have extrapolated, and which is at the same time a real caricature of a tool unethical, a terribly poorly designed artificial system11 :
Meteorologist Eric Holthaus stresses that such a trajectory is unsustainable: by February 2020, he asserts, the Bitcoin network alone could “use as much electricity as the entire world does today.
To the unpredictability of the needs and uses created by new technical solutions is added the lightning speed of their spread in the digital fabric which leads to the sudden and massive increase, by eruptions, of consumption. Once the optimizations have been carried out (boosted “à la Gartner”), there remains a “plus”, a notch, absorbable only by an impossible sobriety in the long term. An established satisfied need and its associated consumption are almost impossible to undo (except by a major, unpredictable and violent change).
This delusion of a “green” digital world
In the meantime, by 2030, there are still many optimizations to be made on such a new and artificial “object”: neural chips, means of communication (a WIFI box, for example, sprinkles all the space while being generally uses from a single point at a time), multiple data center optimizations (by the way, the search for their energy sobriety favors very large data centers and therefore, once again, big players), etc. So, there is work to be done, new technologies to discover, startups to imagine, an economic enthusiasm to share… The “green IT” driver has a bright future.
However, we can rightly consider the digital environment as a whole as a new world “without conscience or limits“. It is impossible to extrapolate what it will be in the long term and in particular its energy consumption. We can at most suspect that the internal logic specific to our technical artefacts rather pushes for an increasing consumption in the long term and only limited by the geophysics, therefore basically unsustainable.
1. ↑ This partly explains our indifference to the subject. We cannot have the physical feeling of the energy consumed, unlike the tank that we fill, the refrigerator that cools the skin, the chimney that spits, the noise of the take-off of an airliner …
2. ↑ Gartner Newsroom – 26 avril 2007 – Press release (broken link)
3. ↑ gesi.org – Mission and Vision
4. ↑ The Climate Group – June 19, 2008 – SMART2020 : Enabling the low carbon economy in the information age
5. ↑ 1 GtCO2e = 1 billion tons of CO2 equivalent, i.e. gas being equivalent to CO2 in terms of radiative forcing.
6. ↑ Simon Roger and Stéphane Foucart / Le Monde – November 13, 2017 – Les émissions mondiales de CO2 repartent à la hausse
7. ↑ Erol Gelenbe, Yves Caseau / Ubiquity (ACM) – Volume 2015, Number June (2015), Pages 1-15 – The impact of information technology on energy consumption and carbon emissions
8. ↑ 1 TWh = 1 terawatt hour = 1 trillion kilowatt hours (kWh). 1 kWh is the energy corresponding to a power of 1 kilowatt developed for one hour. For example, a 20 watt lamp consumes 20 watt hours in one hour, or 0.02 kWh. 1 TWh therefore makes it possible to light 5,700,000 20 watt lamps continuously for a year, etc.
9. ↑ The total electricity consumption of Google, for example, is estimated at 2 TWh, of which only 5% for its search engine. A single request consumes around 0.3 Wh.
10. ↑ Frédéric Bordage / greenIT.fr – February 20, 2018 – La transition numérique fera-t-elle exploser notre consommation d’énergie ?
11. ↑ Jan McGirk / chinadialogue – April 26, 2018 – Is blockchain energy use sustainable?