Why the digital world has a physical footprint
Digital services feel weightless. A message, a video stream or a search seems to exist nowhere in particular. In practice every one of them depends on physical equipment: phones and laptops, home routers, cables and radio towers, and large buildings full of servers. Making, powering, cooling and eventually discarding all of that equipment uses energy and materials.
The usual shorthand for this is a digital carbon footprint: the greenhouse gas emissions linked to digital activity. It is a useful idea, but it is also a slippery one. Published estimates differ, sometimes by a wide margin, because researchers draw the boundary in different places and rely on different assumptions. This article describes the main parts of the picture and is deliberately cautious about numbers.
Three stages: making, using and discarding
A helpful way to organize the topic is by life stage.
Manufacturing. Building a phone, laptop or server involves mining and refining metals, producing semiconductor chips in highly controlled factories, and assembling many components, often across several countries. Chip fabrication in particular is energy- and water-intensive. The emissions released before a device is ever switched on are often called embodied or embedded emissions.
Use. Once a device is running, it draws electricity. For a phone this is small per day, but it adds up across billions of devices. Networks and servers also draw power around the clock. The emissions from use depend heavily on how the electricity is generated, so the same activity can have a very different footprint in places with different energy mixes.
End of life. Devices eventually stop being used. What happens next, including reuse, repair, recycling or disposal, affects how much material is recovered and how much becomes waste. This stage is covered in more detail in our look at electronic waste.
For small, short-lived consumer devices, researchers often conclude that manufacturing accounts for a large share of lifetime emissions. For equipment that runs constantly, such as servers, the use stage is typically more significant. Neither statement holds for every product, and you should treat them as general tendencies.
The main parts of the system
User devices
Phones, tablets, laptops, televisions, game consoles and smart speakers are the most visible part of the system, and they are numerous. Because individual devices are replaced fairly often, their manufacturing footprint is spread across a relatively short working life. That is one reason the debate over how long products last, including the right to repair, is part of the environmental conversation.
The Internet of Things adds many more small connected objects, from sensors to appliances. Each one uses little energy, but the number of them and the materials they contain are part of the picture.
Networks
Data travels through home and office equipment, fixed-line networks, mobile networks and undersea and underground cables. Network equipment is on all the time, so its electricity use is steady. Mobile technologies such as 5G networks are often described as more efficient per unit of data than their predecessors, while also encouraging more total data use. That tension, where efficiency gains meet growing demand, appears throughout this topic.
Data centers
Large facilities host the servers behind websites, apps, storage and cloud computing basics services. They need power for the computers and for cooling, as well as backup systems for reliability. Our explainer on data centers covers what is inside them and why they use so much energy. Data centers are often the first thing people think of when they discuss the digital footprint, but they are only one component, and in some estimates not the largest one.
Emerging workloads
Newer forms of computing draw particular attention. Training and running generative AI systems can require a great deal of specialized hardware, which is discussed in the article on AI energy use. Cryptocurrency mining and big data processing have also been scrutinized. How much these workloads matter overall, compared with everyday streaming, storage and device use, is still debated.
Why estimates disagree
If you read several reports on the digital carbon footprint, you will probably notice different answers. Some reasons:
- Boundaries differ. One study may count only data centers and networks. Another may add manufacturing of consumer devices. A third may include the factories that build the factories.
- Electricity sources differ. The same kilowatt-hour has a different emissions impact depending on whether it comes from fossil fuels, nuclear, wind, solar or hydropower, and this changes over time and place.
- Data is incomplete. Operators of large facilities do not always publish detailed figures, and many devices are in private hands. Analysts fill the gaps with models.
- Usage is hard to attribute. Deciding how much of a shared network’s energy belongs to one video stream involves assumptions. Different but reasonable methods give different results.
- Rebound effects. When technology becomes more efficient, people may use more of it. Efficiency gains do not automatically translate into lower total consumption.
Because of this, it is wise to be skeptical of any single, precise-sounding figure for what an email, a search or an hour of video “costs” in carbon. Such numbers can be a rough guide, but they depend on many assumptions.
Digital technology as both problem and tool
Technology also helps reduce environmental impact in other sectors. Examples include smart grids that help integrate renewable power, precision agriculture that can reduce the inputs used on a farm, digital twins that let engineers test designs without building them, and predictive maintenance that can help equipment last longer. Sometimes these gains are real and measurable. Other times they are promised but not delivered, or are offset by increased use elsewhere. A balanced view recognizes both sides and asks for evidence in each specific case.
What organizations and individuals can do
Without claiming any single step will transform the total, some commonly discussed approaches include:
- Keeping devices in use for longer, and choosing repair or refurbishment where practical.
- Passing working equipment on to others rather than discarding it.
- Using recycling and take-back programs that handle electronics properly.
- Reducing unnecessary data storage and processing, a theme that overlaps with data minimization and data retention.
- Asking suppliers how efficiently their facilities operate and what energy they use, which our article on green IT explores for organizations.
For individuals, the largest effects are often the ones that happen only occasionally, such as how often a phone or laptop is replaced, rather than daily habits like sending messages. Even so, individual choices are small compared with decisions made by manufacturers, grid operators and policymakers.
Questions worth keeping in mind
When you meet a claim about technology and the environment, a few questions help:
- What is included in the calculation, and what is left out?
- Is the figure about manufacturing, use or both?
- What electricity mix does it assume?
- Does it account for people using more of a service as it gets cheaper?
- Who produced the estimate, and can the method be inspected?
The environmental footprint of digital technology is real, but it is not fixed and it is not simple. Honest estimates come with ranges and caveats, and those caveats are part of the information.