This article is part of GP-ORF series — From Alpha Century to Viral World: The Raisina Young Fellows Speak.

---

The world is in a climate crisis, and the predictions for how fast change is coming are getting gloomier. The World Meteorological Organisation’s^<1> most recent annual update on the five-year trajectory for climate change highlights that, until 2024, there is a ~70 percent chance that one or more months will be at least 1.5°C warmer than preindustrial levels and a ~20 percent chance that this will be true for a whole year^<2>. In other words, there is increasing evidence that we will be hitting the tipping point for irreversible change much sooner than anticipated and accounted for in the 2016 Paris Agreement^<3>.

Until 2024, there is a ~70 percent chance that one or more months will be at least 1.5°C warmer than preindustrial levels and a ~20 percent chance that this will be true for a whole year.

To prevent the worst, we need a transformation, and technology has a role to play. Most climate projections stipulate that even with the most ambitious policy and industry efforts, we will be very hard-pressed to stay within the liveable limits of 1.5°C of global heating. The Global Carbon Project graph captures the scope of this challenge by visualising how steep the curve of mitigation needs to be now (see Figure 1). Mitigation efforts starting in 2020 must drop from 40 GtCO₂ to less than half in only five years. It is a monumental transformation. Figure 1: Mitigation curves for 1.5°C Source: Global Carbon Project/Robbie Andrew (CC0 BY 4.0) Most projections refer to “new technologies” or “technological breakthroughs” as part of their assumptions. In some cases, the emphasis is on carbon capture and storage (CCS). In others, the options are left more vague^<4>, acknowledging that: a) we have yet to develop CCS at scale and b) we do not yet know the impacts of CCS tech. Some studies indicate that it may come with unwanted environmental repercussions, similar to negative repercussions of techniques such as fracking^<5>. So yes, we need technology. But we also need more research into the consequences and long-term sustainability of relevant technologies to avoid fixing fatal wounds with Band-Aids.

The internet’s role

It is particularly striking that the one technology that often underlies and powers the research and many of the solutions, has critically undiscussed environmental impacts — the internet.

Sustainability is the interconnection of three elements — social connection, economic wellbeing, and a healthy environment.

The COVID-19 pandemic has yet again highlighted the potential and the necessity of the internet as a fundamental technology in today’s world. Sustainability is the interconnection of three elements — social connection, economic wellbeing, and a healthy environment. From that lens, we have been reminded that the internet is often a lifeline for social connection, certainly in times of physical distancing. It is the primary means to conduct business for those of us that are not on the essential frontlines, allowing many people to work remotely, provide online services, and monetisation. Hence, the internet is a critical vehicle to safeguard economic wellbeing. However, to be sustainable, the internet also needs to assess, mitigate, and live up to its responsibilities for a healthy environment — an element of the equation that is too often neglected. What is the internet’s environmental impact and what would it take for it to be sustainable? Measuring and comparing results from different studies on the internet’s environmental impact is tricky and builds on a range of varying technical assumptions. Here are some facts and figures to assess the internet’s carbon footprint.

To be sustainable, the internet also needs to assess, mitigate, and live up to its responsibilities for a healthy environment — an element of the equation that is too often neglected.

Online advertising

The average internet user of 2019 was served 1,700 banner adverts per month^<6>. With an estimated 4.6 billion people online^<7>, there were approximately 8 trillion banner ads displayed each month. In 2018, a study^<8> evaluated that the carbon footprint of online advertising constituted 10 percent of the total CO₂ emissions of the internet, looking at 2016. In 2016, this meant roughly 106.59 Terra Watthour of energy, which equals approximately 60.28 million metric tons of CO₂ emissions (MtCO₂e). These numbers are likely to have grown since then. Yet, even with this conservative estimate, this would mean 301 MtCO₂e from 2016 to 2020 and almost 1 GtCO₂e until 2030 (out of the remaining global carbon budget of 307 GtCO2e^<9>) for online advertising alone. Research^<10> shows that online advertising only has a success rate of one purchase per million ad impressions. Is it a worthwhile investment?

Connected devices

There are currently around 30 billion devices connected to the internet^<11>, or about three to four devices per person across the global population. Geographically, it equates to roughly 13 devices per person in North America and one to three devices per person in Asia and Africa. In addition, another 14 billion devices through machine-to-machine connections^<12>, like connected cars or smart meters, must also be accounted for. These devices also contribute to the internet’s environmental impact. According to the 2017 Greenpeace Clicking Green report^<13>, the information and communication technology sector was estimated to consume around 12 percent of global electricity demands, with connected devices making up between 30 percent and 40 percent of that share.

Most devices contain rare earth elements, which are mined under questionable circumstances and are difficult to dispose sustainably.

Even socks can be connected to the internet^<14>, checking your body temperature and pausing your streaming experience when you fall asleep. Video streaming, a whole category in and of itself, is estimated to amount to nearly a million minutes of video content every second by 2021^<15>. Or, as the 2017 Greenpeace Clicking Green report notes: “Video streaming is a tremendous driver of data demand, with 63% of global internet traffic in 2015, and is projected to reach 80% by 2020”^<16>. YouTube is also in the running for most-used service globally^<17>. Even if the improved energy efficiency of data centres and processing is slowing the growth of online streaming’s environmental impact^<18>, its overall consequences cannot be ignored. Apart from energy use and contribution to data traffic (data being collected, processed, analysed and stored), all of these devices need to be assessed with a view to their life cycles, including manufacturing, production, distribution and disposal^<19>. Most devices contain rare earth elements^<20>, which are mined under questionable circumstances and are difficult to dispose sustainably.

Artificial intelligence

While often used as a catch-all phrase for automation, we only need to look at a few examples of machine-learning to highlight that the current implementation of artificial intelligence (AI) has a significant environmental cost. Research has shown that training popular natural language processing AI models^<21> produced the same CO₂ as flying roughly 300 times between Munich, Germany, and Accra in Ghana. One of the models looked at is called GPT-2^<22>. In June 2020, OpenAI released GPT-3^<23>, a model that is exponentially bigger than its predecessor. GPT-3 builds on 175 billion parameters, compared to the 2019 GPT-2 model, based on 1.5 billion parameters. Models like GPT-2 were estimated to require 626,155 pounds of CO₂, which equals 284 mtCO₂e or 704,762 miles driven in an average car^<24>. We are yet to calculate the equivalent for the newer GPT-3.

Data centres occupy extensive surfaces of land and put significant strain on global water resources, a factor that is not consistently reflected in the sustainability reports of big service providers.

Moreover, when assessing these huge data processing capacities, we also have to consider data centres, “the factories of the digital age”^<25>, that consume around 2 percent of electricity worldwide^<26>. Data centres occupy extensive surfaces of land and put significant strain on global water resources^<27>, a factor that is not consistently reflected in the sustainability reports of big service providers. As Rob Toews put it in a recent Forbes piece^<28>:

“The “bigger is better” ethos that currently dominates the AI research agenda threatens to inflict major environmental damage in the years ahead. Thoughtful, bold change is needed to set the field of artificial intelligence on a more sustainable and productive trajectory.”

Whether that means investing in tiny AI^<29>, local implementations or reassessing cost-benefits with an eye towards sustainability, there is tremendous room for improvement.

Additional sources of greenhouse gas emissions

The story does not end with online advertising, connected devices and AI. The internet is also grappling with legacy code, spam, increasing website weight, data mining and proof-of-work blockchain implementations, or simply an abundance of useless and outdated data. Without too much detail on these, consider that in 2019, approximately 158 billion spam emails were sent, which amounts to roughly 55 percent^<30> of all email traffic and an average of 0.3g CO₂e per message^<31>. Fake social media accounts add to the size of the problem; for example, in 2019 Facebook removed 2.2 billion fake accounts in the first quarter of the year alone^<32>. Each active profile is estimated to account for 281 grams of CO₂e^<33>. And while different implementations of blockchain may have potential, the technology is mostly known as the system powering Bitcoin and other digital cryptocurrencies. This proof-of-work implementation requires huge amounts of electricity and in 2019, researchers from the Technical University of Munich and Massachusetts Institute of Technology estimated Bitcoin’s emissions at 22-22.9 MtCO₂e per year^<34>.

The internet is also grappling with legacy code, spam, increasing website weight, data mining and proof-of-work blockchain implementations, or simply an abundance of useless and outdated data.

It is abundantly clear that we need a reset. We need to slow emissions and contribute to the sort of transformation required to stay within the liveable limits of global heating.

Different future is possible

If you are feeling overwhelmed, you are not alone. We are learning that sending signals to the market that actively push for better protections and sustainable alternatives is really hard if you cannot picture what such a future will look like. We need images. We need a vision. We need creativity. And a new trend. Imagine this:^<35>

We are in 2050. Welcome to the Museum of the Fossilised Internet. Opened in 2050, it looks back at the obscure practices of the internet of 2020. Gasp at the horrors of surveillance capitalism. Nod knowingly at the plague of spam. Be baffled at the size of AI training data or the reckless excess of data-mining and lament the binge culture of video streaming. These practices are now outdated.

Looking at the main carbon emitters of the internet of 2020 (some captured above) can help us imagine a better world. Visualising a utopia is the first step to making smart decisions to push for change. We must address each of these to help tackle the climate crisis. But we do not have to tackle all of them alone or by ourselves. It is a collective effort, one in which everyone can have a positive impact and make the necessary change.

Visualising a utopia is the first step to making smart decisions to push for change.

Going forward

At this point, carbon capture and storage, or similar alternatives, will be necessary to make up for lost time in reduction and mitigation efforts; however, they are not a panacea. We still need to do our homework and change some habits. Greenhouse gas emissions assessments^<36> should be common practice. These should feed into efforts such as the global carbon tracker^<37> to increase transparency and accountability for everyone. If the idea of the global transformation of societies and markets seems daunting, break it down into little steps^<38> and understand your actions as part of a collective^<39>. Switch energy providers, download instead of stream, repair rather than buy new devices, send fewer emails^<40> — everyone must do this together. It is a commitment, not a one-off. Increasing awareness must be a constant effort, for instance, through tools like Carbonalyser. You can also support your city to take the lead^<41> where national governments may be lagging behind, or engage your co-workers to nudge your company in the right direction^<42>. Ultimately, when looking at tools that most of us use every day, like the internet, we need to take a step back and consider their larger impact. It is about changing our mindsets^<43>, and then meeting people where they are and bringing them along.

---

Endnotes

“New Climate Predictions Assess Global Temperatures in Coming Five Years,” World Meteorological Organization, 8 July 2020. ^<2> WMO Lead Centre for Annual-to-Decadal Climate Prediction, WMO Global Annual to Decadal Climate Update for 2020–2024, WMO, 2020. ^<3> United Nations Framework Convention on Climate Change, The Paris Agreement, UNFCCC, 2017. ^<4> Emily Cox, “Guest Post: How Public Attitudes towards ‘CO2 Removal’ Differ in the UK and US,” Carbon Brief, 7 July 2020. ^<5> Nunez Christina, “How Has Fracking Changed Our Future?” National Geographic, March 2013. ^<6> Dautovic Goran, “The 45 Most Important Advertising Statistics of 2020,” Smallbizgenius, 27 January 2021. ^<7> “Global Digital Population as of January 2021,” Statista, 2021. ^<8> M Pärssinen et al., “Environmental Impact Assessment of Online Advertising,” Environmental Impact Assessment Review 73 (1 November 2018): 177–200. ^<9> “That’s How Fast the Carbon Clock Is Ticking,” Mercator Research Institute on Global Commons and Climate Change, 2018. Brian Dalessandro et al., “Evaluating and Optimizing Online Advertising: Forget the Click, but There Are Good Proxies,” Big Data 3, no. 2 (18 June 2015): 90–102. “Internet of Things (IoT) Connected Devices Installed Base Worldwide from 2015 to 2025,” Statista, 27 November 2016. “Cisco Annual Internet Report (2018–2023) White Paper,” Cisco, 9 March 2020. “Clicking Clean,” Greenpeace International, 10 January 2017. Netflix, “Netflix Make It,” Netflix, 2020. “Video Will Account for an Overwhelming Majority of Internet Traffic by 2021,” Business Insider, 12 June 2017. “Clicking Clean”. Cam Cullen, “Global Internet Phenomena COVID-19 Spotlight: YouTube Is the #1 Global Application,” Sandvine, 6 May 2020. Kamiya George, “Factcheck: What Is the Carbon Footprint of Streaming Video on Netflix?” Carbon Brief, 25 February 2020. “Guide to Greener Electronics 2017,” Greenpeace, 17 October 2017. ^<20> Roger Turner, “Choosing a Better High-Tech Future,” The Science History Institute, 10 December 2019. ^<21> Karen Hao, “Training a Single AI Model Can Emit as Much Carbon as Five Cars in Their Lifetimes” MIT Technology Review, 6 June 2019. ^<22> Emma Strubell, Ananya Ganesh and Andrew McCallum, “Energy and Policy Considerations for Deep Learning in NLP,” Cornell University, 5 June 2019. ^<23> Tom B. Brown et al., “Language Models Are Few-Shot Learners,” Cornell University, 22 July 2020. ^<24> U.S. Environmental Protection Agency, “Greenhouse Gas Equivalencies Calculator,” U.S. Government. ^<25> “Clicking Clean” ^<26> Naomi Xu Elegant, “The Internet Cloud’s Dirty Secret,” Fortune, 18 September 2019. ^<27> Nikitha Sattiraju, “The Secret Cost of Google’s Data Centers: Billions of Gallons of Water to Cool Servers,” Time, 2 April 2020. ^<28> Rob Toews, “Deep Learning’s Carbon Emissions Problem,” Forbes, 17 June 2020. ^<29> Karen Hao, “Tiny AI Models Could Supercharge Autocorrect and Voice Assistants on Your Phone,” MIT Technology Review, 4 October 2020. ^<30> “Global Spam Volume as Percentage of Total E-Mail Traffic from January 2014 to September 2020, by Month,” Statista, 2021. ^<31> Hayley Tsukayama, “How Bad Is Email for the Environment?” Independent, 17 February 2021. ^<32> Emily Stewart, “Facebook has taken down billions of fake accounts, but the problem is still getting worse,” Vox, 23 May 2019. ^<33> Taaj Francois, “How Much Electrical Energy Does Facebook and All Its Users Consume Each Day?” Quora, 10 December 2017. ^<34> Christian Stoll, Lena Klaaßen and Ulrich Gallersdorfer, “Bitcoin Causing CO2 Emissions Comparable to Hamburg,” Technical University of Munich, 13 June 2019. ^<35> MozillaWiki, “Sustainability,” MozillaWiki. ^<36> Greenhouse Gas Protocol, “Greenhouse Gas Protocol,” WRI and WBCSD. ^<37> David Roberts, “The Entire World’s Carbon Emissions Will Finally Be Trackable in Real Time,” Vox, 16 July 2020. ^<38> Erika Reinhardt, “A Data-Driven Guide to Effective Personal Climate Action,” Erika Reinhardt Blog, 28 January 2021. ^<39> Mary Annaïse Heglar, “We Can’t Tackle Climate Change Without You,” Wired, 4 January 2020. ^<40> Sarah Griffiths, “Why Your Internet Habits Are Not as Clean as You Think,” BBC, 6 March 2020. ^<41> Mark Turner, “Fighting Climate Change Starts in the Cities,” International Policy Digest, 10 February 2020. ^<42> “A Guide for Employees: How to Make Your Tech Company Sustainable,” Medium, 3 August 2017. ^<43> Cathleen Berger, “Sustainability Needs Culture Change. Introducing Environmental Champions,” Mozilla, 13 July 2020.

• Media and Internet
• Internet Governance
• Sustainable Development
• America
• Bhutan
• The Pacific
• East and Southeast Asia
• East Asia
• data
• Internet
• Legacy Code
• Rare Earth
• Social Connection

The views expressed above belong to the author(s). ORF research and analyses now available on Telegram! Click here to access our curated content — blogs, longforms and interviews.