Digital Carbon Footprint — What can we do ?

Disclaimer: I am by no means an expert in the field of energy production, green IT or related subject. The following information is just the result of personal research.

Everybody tackles the subject of environmental issues their own way. Some think there is no issue, some believe it is an issue that humans have no impact on, and most of the others agree that the human race has effect on the earth ecosystem that has negative impact on life (whether plant, animal or human based).

Although I’ve always been on the latter side of the spectrum, the problem was always kind of distant to my life for 3 main reasons that I think are shared by many:

1. I believed the biggest problems would have technical solutions that experts would come up with and save the day
2. I thought these issues should be handled by the governments or large and influential associations
3. I also had the belief that I could not have real impact at my scale

After watching a lot of Youtube channels and reading a fair share of medium articles about economics, science and ecology, I realized that most of my previous assumptions where biased and deserved a second look at.

As environmental issues spread a large spectrum, I’ll focus on the one I’m professionally the closest to : energy consumption of the digital industry and it’s impact on the world.

The state of the digital World footprint

IT global industry amounted to an approximate 10% of the world’s global energy in 2013. Moreover, the increase of the sector’s consumption is estimated to 10% per year, which is more than the expected increase of the produced renewable energy.

If IT was a country, it would be the third biggest energy consumer in the world in 2012:

Data centers consume 3% of all globally generated power at run time (which means the power necessary to run the machines, compared to the build time — power necessary to build the machines). Even though they are getting more and more efficient thanks to AI driven optimization and the Green Data Center movement, it can be somewhat ironic that these data center are more and more used to train ever greedier machine learning algorithms (which can emit as much as a round trip trans-America flight of equivalent carbon dioxide — CO2 eq is a measure of the climate change effect of greenhouse gas emitting systems)

Even though I have not been able to find sources about the distribution of the energy consumption of the IT industry, I list here what I think are the main categories :

• The internet network : used mainly to power end consumers demands, but industry and IoT (Internet of Things) is already a big consumer and expected to grow exponentially
• Cloud computing and hosting: data-centers are used everywhere to power B2B and B2C software, SaaS products and run AI algorithms. Some of them consume more power than 180 000 US homes.
• The end consumer’s devices: laptops, workstations and smart devices. For instance, if we take a typical Dell computer that emits around 200kg CO2eq to be made and delivered (not the run time consumption), and multiply it by the number of laptops manufactured a year (around 300M), we have 60 Mt CO2eq, which is more than the total yearly emission of Niger in 2017
• Internet software providers, especially media providers: Youtube is estimated to emit 10 Mt CO2e , which amounts to roughly a third of the internet emissions. Youtube and Netflix combined represent at peak more than half of the data usage in America.
• Radio based and satellite mobile network are still a big part of the energy consumption of the ICT (information and communication technologies), but it is the digital infrastructure that increases the most, not the radio network construction and operation

Most of the cloud companies really commit on renewable energy, with Apple and Google leading the race by now being 100% carbon neutral (which does not mean they run on 100% renewable energy, just that they buy as much renewable energy as they use).

It is interesting to see that AWS (Amazon web services) only reach 50% of renewable energy buying in 2018, even though they are the oldest and biggest of all cloud providers. This clearly means that it only depends on the company’s conviction and mindset to adopt green energy. This might be nuanced by the fact that their huge fleet of servers and seniority — thus technical debt — could make the transition harder than for following competitors.

Chinese digital companies (Alibaba and Tencent being the biggest) for now do not share this mindset, as they do not communicate much about their clean energy efforts and it is estimated that they use coal a lot (2/3 of their energy use according to the Greenpeace report of 2017). That is too bad because they are the fastest growing digital industries in the world (86.1% in 2018, with AWS having “only” grown by 48% in the same year).

Evolution of the IT consumption

The energy consumption of the IT industry is likely to increase drastically in the coming years. The SMARTer 2020 report estimates a 81% increase in electricity demand of data centers in 2020 compared to 2017. This is a higher percentage than the globally estimated 60% increase of electricity demand in the same time frame all sectors combined.

There is 4 main factors that drives the energy demand of the industry :

• The market is ever expanding with new users from developing countries coming in masses. For instance, Chinese internet users went from 500 millions in 2012 to 800 millions in 2018
• The connected users spend more and more time using their smart devices (thus consuming energy to power apps and software)
• New highly evolved usage of smart device (prediction, assistance, …) requires new algorithms and connectedness between devices, thus increasing demand for computation and network energy
• GDP (Growth Domestic Product) generated by ever power demanding high tech products increases faster than the optimizations in energy consumption

There is also, in my opinion, an economic side of things that makes the Internet more and more energy greedy. Indeed, the main business model of the IT industry is advertising (99% of Facebook’s revenue and 87% of Google’s revenue). And advertising is very energy greedy (at least 10% of all the IT energy consumption).

In terms of physical machines, the number of servers running in data centers keeps increasing exponentially:

• AWS number of servers estimate from 158 000 to 450 000 in 2013 to 1.3M in 2018
• Microsoft went from 1M servers in 2013 to 3M in 2018
• Google went from 900 000 servers in 2011 to 2.5M in 2016
• Alibaba cloud went from 25 000 servers in 2013 to 450 000 servers in 2017 (web facing servers only, probably around 3 times that adding the not web serving ones)

Here is a diagram of these number to better see the increase rate :

Is there a problem ?

So, here are the main key points we can pull out for all these data:

1. First, the digital industry worldwide keeps consuming more and more energy at run time
2. The biggest players that run the internet and the cloud, although optimizing their consumption thanks to engineering and machine learning, keep demanding more energy to match the usage growth
3. Even if all the digital industry actors (especially China) committed equally to being carbon neutral, the renewable energy production probably could not keep up with the energy demands for now
4. At build time, the industry keeps producing more and more machines (servers, captors, optic fibers, smart devices…) that also come up with a huge carbon footprint
5. All of the former facts are mostly driven by the ever growing created need for more entertainment, efficiency, and social connection in the consumer’s everyday life.

So we can already see that the problem of the energy transition is somehow paradoxical: besides struggling to get “green energy” wildly adopted by major industry players, its production as we know it now could not keep up with the increasing demands.
Another paradox can be found in the rebound effect (also called Jevon’s Paradox): as cloud computing is getting more and more efficient, it’s usage is accordingly spreading, new actors come in as the prices drop to take advantage of the scale of the industry. It could even have unexpected side-effect that would go beyond the expected snowball effect.

This is about to get even more complicated with the introduction of rare earth and its impact on the renewable energy business.

Introducing rare earths

To explain why the former key statistics might pose problems, we should first talk about rare earths. There is a great book for french readers by Guillaume Pitron tackling this subject.

Rare earths are a group of chemical elements that have specific magnetic and chemical properties that make them particularly useful (even essential) to build technical products.

Their specific properties enable the production of magnets that are used for motors (for hybrid or electric vehicles but also wind turbines or all the small motors you have on your oil powered car — windshields, mirrors, …), but also electronics (hard drives, smart devices, everything having a screen, a circuit board or a magnet).

I won’t go into details on the use of rare earths in technology, but basically, they are everywhere. Without them, we wouldn’t have current day Internet, cloud providers, industrial machines, smartphones and computers, and also renewable energy.

Despite what their name suggest, rare earths are not actually that rare. They are present all around the globe in low concentration, and new fields are discovered regularly.

The main issue with rare earths is the extraction, which is very costly and not environment friendly to say the least. Mineral extractions requires the usage of a lot of chemicals that if not properly handled can end up polluting soils and water. There is also risks of radioactive tailing expelled from elements ores.

There is also the geopolitical aspect of rare earth production that is problematic. As you can see in the following chart, China amounts to around 80% of the global rare earth production, even though they only possess around a third of the deposits:

The USA used to extract rare earth, especially in Mountain Pass, but the mine closed do to environmental restriction following many lawsuit regarding repeated radioactive leaks, and also because of aggressive competition from Chinese suppliers.

Basically, what happened was that China realized they were sitting on top of a pile of soon to be essential raw materials, and they implicitly agreed to decrease environmental regulations to extract huge amount of rare earth to meet market needs. It was a win-win situation : western countries could import cheaper rare earth to produce high end products that they labelled “green” and promoted as “dematerialization”, while China made a lot of money exporting rare earth or primary electronic materials that uses rare earths.

That lead to the situation we have today:

• China faces huge environmental and health problem due to unmonitored rare earth extraction: the extreme cases of cancer villages express how bad the situation is
• On the economic side, Chinese administration realized it would be far more profitable in term of money and international recognition to export high end products rather than raw materials. They decided to impose quotas on exportation to ensure their industry would get the most of their rare earth extractions.
• They also increased environmental regulation that implied production impacts and increase of costs
• The western countries hypocritically tried to get the WTO (World Trade Organization) to prevent the quotas so that China could keep being the pollution garbage of the world, but were unsuccessful

To summarize, the growth of digitalization and the advent of the so-called “Energy Transition” that is so cherished by tech entrepreneurs depends on the production of materials which have the following characteristics:

• We have no control over the market
• We have no idea how to extract at the same rate without huge environmental and health impacts
• Although the materials are not scarce, they are limited enough so that we can’t replace fossil fuel energy with renewable ones.
• We also find it extremely difficult to recycle, mainly because of the diverse and small mixtures of those elements in technology products

To me, this list of “known unknowns” and technical challenges makes my first assumption that experts would come up with technical solutions harder to believe, or at least yet to be proven.

The second assumption, that is the role of institutions and governments to tackle these issues is a bit trickier, as they are indeed chasing the problems and fixing some of them with green taxes for instance.

But to me the main issue, which is how to produce real green energy at the accelerating pace of technologies is not handled and discussed enough by the authorities, whether consciously or unconsciously. Crippled by the blind faith in an ideology that set GDP (Growth Domestic Product) as the sole key performance indicator for the progress of the world, institutions’ answer to climate change and pollution will never do anything that might lead to GDP stagnation or decrease.

Yet GDP is highly correlated to energy consumption, some research even suggesting a causality, meaning that there can be GDP growth only if there is more energy consumed.

But the fact is, to keep increasing the energy usage of our world is very hazardous for the environment, health, and the fate of life on Earth. So the safest solution would be to try and design a new way to outlook growth that does not imply increasing energy supply, until hypothetical new technologies or unicorn optimization algorithm keep the energy consumption under control.

But the mental shift required to do so is a big leap that would put the system and the people who most benefit from it at risk, and it is unlikely that a change will come from within the system itself.

What can we do ?

As said before, the ICT business is very much driven by the mass consumption of new and attractive technologies. So the biggest impact one can have on the carbon footprint of these technologies is to cherry pick them according to their impact and on average use less of the same technologies.

As a digital products consumer:

The first step towards reducing your carbon footprint would be to reduce the amount of new electronic device you build. Take smartphones for instance : 85% to 95% of their carbon footprint comes from the building of it, the rest is the energy consumption usage. So instead of buying the shiniest smartphone every year, consider keeping it as long as you can, and if you really must change, opt for a refurbished one (Backmarket is doing a pretty good job labeling and reselling refurbished phones).

However, the usage energy consumption of smartphones is between 10 to 100 times less than the laptop one. This is mainly due to the larger screen. So if you have to choose between using, say Google Maps on your laptop or phone to get directions, you can pick the smartphone without hesitation. Of course, shutdown down the devices you do not use instead of putting them is sleep mode is significant and requires little effort.

As a digital services consumer:

You can try to favor the use of services that exchange the least data information with your device, keeping in mind that:

• Video (download, streaming, buffering) bears a lot of data, and displaying it requires lots of compute (thus energy) from your device
• Audio is next in line
• Images can be optimized but still represent lots of data
• Text is the least consuming of the main content types

There are a lot of behavior that can change without impacting the user final experience. For instance, if you want to listen to music, prefer the use of a music cloud service, or buy the song online instead of playing it on YouTube if you won’t even look at the video. If your goal is to communicate with someone for a place to meet for instance, prefer send a simple text instead of sending a selfie, audio message or video calling them. A text message is around 200 times more efficient than an email for instance.

Some cloud services in my opinion do not provide enough added value compared to their energy consumption. For instance, cloud file storage consumes more than a million times more energy than file storage (external hard drive or USB stick for instance). I understand the value of multi-device instant access to personal files and safe backup, but is it worth the impact ?

For all the cloud services that you judge necessary for your personal life, try to use the ones that relies on renewable energy as its main power source. Or at least pick your provider according to their involvement in energy saving. For instance, Dropbox hasn’t setup any actions toward reducing their energy consumption to reduce their environment impact.

Out of curiosity, I extracted the usual carbon footprint as computed by different research. These numbers should be handled cautiously, as they are a large estimation and probably do not embed the whole spectrum of dependencies for each actions.

If you’re a techie person:

There are multiple actions you can take to reduce your energy footprint during the whole development cycle: development and testing, building, running.

During development, you can choose a programming language that does not require high end development environment, that uses less memory and CPU. This report makes various benchmark on energy and memory consumption for the main languages. We can see that C and Rust for instance consumes at least 20 times less energy than Ruby and Python for instance. Of course, that does not mean you should use C for every software, because there is a compromise to find for you between energy efficiency and development time. If it takes 20 times longer for you to develop in C instead of Python, the energy efficiency makes no sense. I now use this benchmark when I am hesitating between 2 languages for a specific use case.

For the continuous integration and hosting of your software, you can have a look at the “cleanest” cloud providers to support your provisioned infrastructure. According to many reports, Google Cloud appears to be the most coherent in its action regarding involvements with clean energy.

Also, unless you have expertise in that area, do not host your software yourself, it’s very likely you will consume more energy than the cloud equivalent.

If you are feeling adventurous, there are specific hosting services that focus on having low impact on the environment : GreenNet for instance, go as far as to have no company vehicles, having to bike to the data centers in case of needed maintenance. Of course because of its smaller scale it won’t have the same prices or reliability as more traditional providers.

Finally, as a citizen:

The biggest action you can take that could have effect at scale still remains the vote. Keep in your mind the limit of energy transition advertised by the main political parties during national and European elections. Try to be aware of the ideas of the different candidates when they mention climate change and renewable energy.

Do they mention the geopolitical and environmental issues with rare earth, and its impact on renewable energy production?
Do they assume the correlation between growth and energy consumption?
Are they aware that increasing energy efficiency does not necessarily imply decrease in consumption (thanks to the rebound effect)?
Do they advocate systems that will increase the industry’s energy consumption?
Do they take for granted the idea that technological progress will resolve every issue regarding climate change ?
Are they open-minded about a change in the way we see growth, or even consider degrowth as a valid topic of discussion, or do they rather follow blindly the liberal dogmas without even considering another model ?

Conclusion

With this article, I aimed at providing you with new outlooks on the challenges of energy consumption and environmental impacts, especially regarding the IT industry.

I don’t want you to think the IT business leads thoughtlessly to the destruction of life on earth. I still believe most of the companies are aware of their impact on the world and try to balance that with good measures and power management optimizations. Thank goodness for the innovative researchers and engineers that manage to reduce the PUE of their data centers to temper with exponential usages, or the telecommunication engineers that reduce the power needed to transfer ever increasing data loads.

I would just want to raise awareness on the fact that the issue is more complicated than just a matter of shifting investment from fossil fuels to green energy.

In my opinion, energy demands must decrease to stop the destruction of ecosystems and the effects on peoples health. To do that, change must also come from within, I mean us the citizens, so we can grasp our own impact on the matter and take actions as product and services consumers but also voters and potential decision makers.