The Internet’s Environmental Footprint. Part 1.
Ever wondered what the internet’s environmental impact is, how to make sense of greenhouse gas emissions assessments, whether climate neutrality is achievable, or also, what it is that we can do to improve and move towards a sustainable future for all?
You’re not alone.
I’ve grappled with these undoubtedly very big, complex questions for a while now and though I’ve occasionally shared research and initial ideas through talks and previous essays, I had yet to write a concise summary. So here it is. In 4 parts, three of which will focus on what we can do going forward. Please consider this an invitation to engage, to think, to ask questions, and to grow together.
Today, I’ll start by sharing some rough calculations to assess the internet’s environmental impact, including social media, online advertising, video streaming, connected devices, and artificial intelligence. I’ll do my best to put these into perspective, offering comparisons for scale and clarification on methodology where possible.
Over the next 2–4 weeks, I will then focus on what needs to happen next if we are to tackle this challenge, looking at regulation, change management in businesses, and narratives that support shifts in mindset respectively.
Ultimately, it is simple: We are in a climate crisis. And by now, you’ve probably heard that most climate projections point out that even with the most ambitious policy and industry efforts, we will be very hard pressed to stay within the livable limits of 1.5°C of global warming. Note: 1.5°C is the same as a 2.7°F temperature increase.
We are still adding layers and layers of greenhouse gases around the planet and even after we’ve stopped emitting more, we will still be suffocating — unless we learn to draw down more emissions than we release into the atmosphere.
At the current rate of somewhere between 40–45 gigatons (Gt) of emissions per year, the remaining carbon budget before the planet becomes uninhabitable for large parts of the global population is scarily low: 276 GtCO2e (gigatons of carbon dioxide emissions equivalent, accessed on June 3, 2021).
This translates to roughly 6 years and 6 months until we surpass our ability to keep global warming below 1.5°C.
So, in case this wasn’t clear: we’re talking about our problem, not some distant future.
Making Sense of Scale
To clarify: 1 gigaton is the same as 1 billion metric tons. And because I know how incredibly difficult it is to understand these numbers and put them into meaningful perspective, I’ll give you a few comparisons.
The average annual emissions per person are 4 metric tons (mt) or 0.000’000’004 Gt. Obviously, emissions are not evenly distributed across the globe. The German average for example is 11 mt per year per person.
One gigaton in weight is also around 200 million elephants — more than the 440.000 that live on the planet. It also equals 3 million Boeing 747 jets, of which 1.573 have in fact been manufactured.
In emissions, it’s a little less than the annual emissions of all of Indonesia with a population of 270 million.
Sustainability and the Internet
According to the European Commission, the impact of the Information and Communications Technology sector amounts to around 2% of global emissions — and it cautions that this may increase to up to 14% until 2040, if no countermeasures are taken. This illustrates both: the growing impact of digital technologies and its outsized opportunity to fuel the necessary transformation.
Even at 2% that would equal roughly 0.8 GtCO2e or 800 million mt emissions per year for the sector. To compare this once more, 0.8 GtCO2e are about the same as Germany’s annual emissions — or, about 160 million elephants in weight.
To make better sense of these assumptions, let’s work through a few approximations. I’ll share some facts and figures but I do want to be clear that all of these are merely back-of-the-envelope calculations that are supposed to help us grasp the scale.
This is not about letting fossil fuel companies and major polluters off the hook. Instead, I want to point to avenues where progress is possible and identify aspects of our digital infrastructures that come with unintended consequences that are simply too severe to look the other way.
Let’s start with Social Media.
I found a 2015 study that looked at Facebook’s reported emissions and concluded that this would amount to 281 grams of CO2e per active profile. While the assumptions in that report are rough to say the least, it does provide some orientation for how to approximate the impact of social media.
The impact per platform will absolutely differ but lacking insight into each of them, we can take these 281 grams and multiply them by users for each of these platforms. And yes, there are more social media platforms out there, I’ve just picked some of the popular ones and used the usage numbers that are available on Statista to get a rough overview.
Taken together, we’re looking at over 5 billion users.
- Facebook: 2.7 billion
- Instagram: 854.5 million
- TikTok: 800 million
- Snapchat: 347.3 million
- Twitter: 321 million
- Twitch: 15 million
And by multiplying 5 billion social media users with 281 grams, this results in roughly 1.4 million mt. Or: According the US Environmental Protection Agency, the amount of energy required to power over 161.000 homes for a year.
Turning to Online Advertising.
One often quoted study evaluated that the carbon footprint of online advertising in 2016 constituted 10% of the total CO2 emissions of the internet. In 2016, this meant roughly 107 Terra Watthour (TWh) of energy, which equals approximately 60 million tons of CO2 emissions (CO2e).
If you’re baffled by the amount, just consider that the average internet user of 2019 was served 1,700 banner adverts per month. With an estimated 4.7 billion people online, there were approximately 8 trillion banner ads displayed each month.
We can argue about whether the market has since grown or not, but either way, this is a huge number for online advertising alone. Again, for perspective, this amount of energy would power almost 7 million U.S. homes for an entire year.
Then there’s Online Streaming.
The measurements and data shared around this category have drawn a lot of attention and are currently quite disputed in terms of emissions.
You may be among the 167 million Netflix subscribers, or you may be part of the users that boost YouTube to be in the running for most used service globally. According to YouTube itself, we’re talking about 1 billion hours of video watched every single day.
Research puts the global average at 36g CO2e per hour of streaming. If we were to multiply this by the 1 billion daily YouTube hours and by 365 to cover the year, this would put us at 1.3 billion mt or 1.3 Gt — and that’s just YouTube, Netflix, Amazon Prima, Disney+, Apple TV and others will add their share. If you’re trying to compare: we’ve reached 260 million elephants in weight or emissions just short of all of Japan’s annual emissions or about a third of the EU-27 annual emissions.
While improved energy efficiency of data centres and processing is slowing the growth of online streaming’s environmental impact, these emissions can hardly be ignored.
How about Connected Devices?
There are currently around 30 billion devices connected to the internet. This amounts to 3–4 devices per person across the global population. Geographically, it equates to roughly 13 devices per person in North America and 1–3 devices per person in Asia and Africa.
Apart from energy use and contribution to data traffic, meaning data being collected, processed, analysed, stored, all of these devices need to be assessed with a view to their lifecycles. This includes manufacturing, production, distribution, and disposal.
It is incredibly hard to find proper assessments for all the connected devices out there, though I’ll mention Apple, which provides insightful reporting. The average lifecycle emissions for an iPhone 11, for example, amount to 75 kg per device. The exact amount depends somewhat on available storage and features. But if we were to multiply the 30 billion connected devices with these 75kg, this adds a staggering 2.25 Gt. Now, whether the lifecycle of a device is 2, 3, 5, or 10 years, will likely vary — meaning emissions per year could range between 1.1 and 0.2 Gt.
Another bucket is Artificial Intelligence.
Artificial Intelligence (AI) is often used as a catch-all phrase for automation and I will admit that I have no idea how many models and implementations really exist. Thanks to research from MIT we know, however, that training popular natural language processing AI models produced the same CO2 as flying roughly 300 times between Munich and Accra, Ghana, namely 284 mt CO2e.
There are countless such models with similar or even bigger scope, so we must assume a multiple of 284 mt, which is the amount required to power some 33 homes for a year.
These back-of-the-envelope calculations get us to at least 1.5 GtCO2e and likely to over 2.6 Gt, which would far exceed the 2% estimate shared by the European Commission, and it may be way off on the lower end still.
Again, Germany’s emissions in 2019 were around 0.8 Gt or 2% of global annual emissions. At 1.5 Gt, we’re looking at almost 4% of global annual emissions, at 2.6 Gt this climbs to around 7% and as you can see, chances are that AI and impacts I may be missing push this estimate even further.
But. It’s also true that the picture is incomplete because I don’t have enough data to balance the costs with the internet’s benefits. That aside, as is, and speaking for myself, I am not comfortable with this scale of the internet’s greenhouse gas emissions.
So, if you’re feeling overwhelmed, you’re not alone. In the upcoming parts of this series, I’ll focus on what we can do to start not just understanding, but improving and mitigating the growing environmental impact of the internet.