Miguel Coma co-authored this article.

In 1801, a French merchant invented punch cards to create a kind of data storage and make a loom weave fabric automatically. Call those punch cards an early computer program. In 1853, a Swedish father and son designed a calculator that could print. In 1941, two men designed a computer that could remember and store information and perform an operation every fifteen seconds.1

Indeed, early computers expanded our abilities. Some were as large as a room. Today, computers fit in our hands. We use them for most aspects of daily life. Children typically use computers before they have speech.

While engineers imagine increasingly fabulous devices and applications, few question the relationship between digital tech and environmental degradation. Few scientists study the ecological impacts of manufacturing, operating and discarding users’ digital devices, network infrastructure and data centers. Impacts include the use of fossil fuels, greenhouse gas emissions, extractions (more than 50 metals), chemical manufacturing, disrupting wildlife habitats on land and in the sea, taking water from communities and farmers, and radioactive and toxic waste.

Who monitors digitalization’s footprint?

To the best of our knowledge, no single standard methodology exists to analyze the (scarce) data we have about these impacts. One study might look at the footprint of manufacturing, operating and discarding digital devices (mobile phones, laptops, e-vehicles)—but ignore smart TVs. Another might focus on electricity use or CO2 emissions of digital tech—but ignore blockchains.

The European Union names 14 categories of environmental impacts2. Of these, CO2 emissions are the most studied. To date, no peer-reviewed study considers the digital industry’s combined ecological impacts.

In 2020, Charlotte Freitag 3 and colleagues published a synthesis of recent peer-reviewed studies about info-communications-technologies’ (ICTs’) environmental impacts, with a focus on CO2 emissions. The researchers report that while “ICT has delivered increasingly wide-ranging efficiency and productivity improvements to the global economy…global CO2 emissions have also risen inexorably.”

Digital tech now accounts for two to four percent of global CO2 emissions (civil aviation is two percent)—or 1.2 to 2.2 gigatons of CO2 equivalents (GtCO2e).

Freitag asserts that reducing digital tech’s CO2 emissions would require major actions from industry, legislators and regulators—and that current “Green Deal” policies cannot achieve sustainable goals alone. We agree. We also recognize that current regulations do not decrease CO2 emissions and mining’s impacts on Indigenous communities or wildlife habitats. They don’t decrease electronic waste. We see that engineering schools rarely cover these issues.

What policies would encourage engineers to design devices and infrastructure that biodegrade and that can be repurposed (given extended use)?

Big tech to save us all?

Some big tech corporations have vowed to reduce their own CO2 emissions. In actuality, these “reductions” may include carbon offsets that do not eliminate emissions. Or, as Freitag reports, they are too weak to reach net zero targets by 2050. Plus, no agency monitors or enforces these commitments.

Meanwhile, three digital technologies generate considerable, unchecked growth in CO2 emissions: Bitcoin cryptocurrency, the Internet of Things (IoT) and artificial intelligence (AI). Unfortunately, without evidence, the European Commission presents these technologies as key levers that reduce emissions. This deceptive narrative fails to recognize climate targets. It also ignores the paradox that increased efficiency creates a rebound effect: it leads to more manufacturing and cheaper devices, which in turn increases energy use, mining, emissions, water use and waste.

New digital trends or threats?

Big data and AI present opportunities and threats to mitigating climate change. For example, they might reduce road traffic congestion. But apps such as Waze 4 can direct traffic congestion without AI—while training one AI to solve problems can generate between 4.5 kg and 284 tons of CO2 (five times a gas-powered car’s lifetime CO2 emissions). AI’s growth between 2012 and 2018, measured in “petaflop/s-days” (the number of computer operations while the machine learns), increased 300,000-fold.

The Internet of Things (IoT) also grows exponentially. By 2025, it could connect 75 billion devices—then grow, annually, by 19%. Most of these devices will support home automation, security and surveillance (48%). The fastest-growing IoT sectors are cars (30% annual growth) and “smart” cities (26%). Note: few researchers study carbon emissions from manufacturing connected devices.

Last but not least, blockchain tech (in particular cryptocurrency) emits as much CO2 as entire nations 5. Bitcoin’s 2020 electricity use could have powered seven million US households. Researchers estimate that bitcoin demands twelve gigawatts (GW) of power—the power generated by seven large nuclear reactors. A single Bitcoin transaction consumes around 750 kWh, enough for a two-ton electric car 6 to drive 5000 km (3100 miles). Moreover, half of Bitcoin transactions happen in China, where electricity runs largely on coal, the fuel that emits the most greenhouse gases.

Rebound effect guaranteed

While efficiency drives electricity demands down, increased traffic drives them up. Data traffic consumes significant amounts of electricity; and video streaming accounts for over 80% of data traffic. New Internet infrastructure such as 5G allows for increased video traffic; because of the rebound effect, 5G increases traffic, electricity use—and CO2 emissions. New, 5G-compatible devices require more mining and manufacturing and generate more e-waste.

Given the limited number of Internet users and the limited number of hours per day, some experts predict that traffic and data demands will plateau. This assumes that machine-to-machine data is negligible. Alas. 5G-connected high-definition video surveillance cameras alone could create exponential growth in data traffic and processing.

Does ICT enable carbon savings in other industries?

Many governments’ “green deals” assume that digitalizing health, education, agriculture, banking, manufacturing, transportation and building will save enough CO2 to counterbalance the digital industry’s emissions. Yet these industries’ 15% total reductions in CO2 emissions (achieved only in ideal conditions) would fail to help us realise climate targets. In fact, because of rebound effects alone, digitalization could increase ICTs’ global CO2 emissions! Reductions will occur only when digital tech eliminates use of carbon-intensive technologies.

Can renewable energy decarbonise ICTs?

The digital industry proudly promotes its increasing use of “renewable” energy. However, industrial wind turbines, solar arrays and batteries are “renewable” and “clean” in name only. Manufacturing these technologies requires extractions and fossil fuels; it generates CO2 and hazardous waste. Then, because wind and solar generate intermittent power, they depend on backup from natural gas or coal (fossil fuels), nuclear power or toxic batteries. Further, they cannot power energy-intensive manufacturing processes like smelting.

What could truly reduce carbon emissions, extractions, water use and toxic waste? To make data centers and network infrastructures carbon neutral, could big tech limit its energy consumption? Could it restrict energy-guzzling (yet non-essential) services like 4K or HD video streaming?

Can current policies help us reach “net-zero emissions”?

Climate policies that promote energy efficiency, renewables and circular electronics to help us reach “net-zero emissions” are not realistic. To begin, researchers estimate that the digital industry’s carbon footprint increases annually between six and ten percent. Adhering to the Paris Agreement (to keep below a 1.5°C global temperature increase) will require digital tech to reduce its CO2 emissions by 42% by 2030, 72% by 2040 and 91% by 2050.

Regardless of how often the claim is repeated, efficiency gains in other sectors will not compensate for the digital industry’s own growing extractivism 7, water use8, waste 9 or emissions.

Freitag suggests that only annual constraints on consumption (i.e., COVID), or a tax or cap on carbon emissions, could actually help reduce CO2 emissions significantly.

In any case, reducing our environmental impacts will require reducing the production and consumption of digital goods.
Who’s ready to sustain a digital diet?


1 History of computers: A brief timeline.
2 2013/179/UE Commission Recommendation of 9 April 2013 on the use of common methods to measure and communicate the life cycle environmental performance of products and organisations. Climate change, Ozone Depletion, Ecotoxicity for aquatic freshwater, Human Toxicity – cancer effects, Human Toxicity – non-cancer effects, Particulate Matter/Respiratory Inorganics, Ionising Radiation – human health effects, Photochemical Ozone Formation, Acidification, Eutrophication – terrestrial, Eutrophication – aquatic, Resource Depletion – water, Resource Depletion – mineral, fossil, Land Transformation.
3 Freitag, Charlotte et al., The climate impact of ICT: A review of estimates, trends and regulations, Physics and Society, 2020.
4 Waze does not seem to report using Artificial Intelligence (AI) for navigation but chose AI for its more recent carpool service to match drivers and riders (a choice not making AI necessary).
5 Bitcoin is by far the most popular cryptocurrency. While writing this article in February 2023, unsustainable cryptocurrencies Bitcoin, Ethereum 1.0, Monero, and a total of 316 currencies use “proof of work” (PoW) consensus algorithm to validate each transaction, consuming high amounts of energy. On the other hand, 243 cryptocurrencies such as Ethereum 2.0, Cardano, and Solana use “proof of stake” (PoS) that requires much less energy. Finally, some currencies use their own consensus mechanism.
6 The Tesla Model 3 weighs 1.8 tons unladen and consumes 151 Wh/km. With 750 kWh of electricity used by one Bitcoin transaction, this car can drive 750,000/151 km or about 5000 km (3100 miles).
7 Sovacool, Benjamin K. et al., “Sustainable minerals and metals for a low-carbon future,” Science, 3 Jan. 2020.
8 Asianometry, “The Semiconductor Water Problem,” 2 Sept., 2021.
9 Lepawsky, Josh, Reassembling Rubbish: Worlding Electronic Waste, MIT Press, 2021.