Your smartphone is an amazing piece of technology. It has the computing power of a desktop computer, lots of on-device storage, fast connections, and a high-end digital camera. You can go for days on a full charge, so it seems that the smartphone is a model of energy efficiency.
It is, but there’s a catch. Most of the tasks you do on your smartphone aren’t actually
processed on the device. Much of it happens in the cloud — in other words, someone else’s
computer. Those cloud computers are powerful, energy-hungry servers housed in data
centers, often in large numbers.
When you back up your data, run an app or game, or execute a search or AI query, most, if
not all, of the processing takes place on those data-center servers. The demand for data-center processing power is fueling rapid data-center expansion.
That expansion is dramatically increasing energy consumption nationwide and placing stress on the electrical grid. It’s raising electric bills and disrupting climate initiatives.
Our online habits for both personal and professional use drive a significant portion of the
demand for more data centers. The energy needed to compose a 100-word email using
ChatGPT, for example, equals that needed to power 14 LED lights for an hour, according to a study by the Washington Post and the University of California.
That means you have some control over the growth rate of data centers. Before I explain how, here’s a little background.
How data centers affect energy demand and electricity rates
The servers that make up a data center use advanced microprocessors and require a lot of
support and data storage systems. They require a vast amount of electricity to run. For
example, a server typically draws 200 to 400 watts while processing video streaming and more than 600 watts while processing AI.
Large-scale data centers have tens of thousands of servers requiring more than 100
megawatts of power. That’s enough to power about 80,000 homes, according to the
U.S. Department of Energy.
Servers and storage make up only about 26% of a data center’s energy consumption,
according to a study by ESDS Software Solution. Cooling places the biggest demand for
electricity at 50% of total consumption; servers generate a lot of heat. Network hardware,
power conversion, and lighting make up the rest of the demand.
A data-center-related energy crisis in some states serves as a warning to the rest of the
country. PJM Interconnection runs the nation’s largest electrical grid, serving 65 million
customers in all or parts of 13 mid-Atlantic states and the District of Columbia. This includes Data Center Alley, the world’s largest data-center cluster, in Virginia. A report by Monitoring Analytics predicts that data centers will add nearly 12,000 megawatts of load to PJM’s grid in 2026. Peak load growth is expected to reach 32,000 megawatts by 2030.
PJM must purchase energy in advance to ensure that it can deliver electricity to meet that
load. In July, costs to meet electricity demand from data centers in the region increased 82% or $7.2 billion. That cost was largely passed on to all ratepayers, not just data-center
operators.
It’s not only the cost of the energy that’s passed on to ratepayers. Bringing data centers and the energy sources needed to power them online comes with interconnection and grid
upgrade costs. From 2022 to 2024, utility companies within PJM’s region started 150
transmission projects devoted solely to connecting data centers to the grid. The increased
energy demands and interconnection costs could increase utility bills by as much as $163
billion within the region, according to research from the Natural Resources Defense Council.
How to Minimize Your Online Energy Footprint
The scale of energy demand growth from data centers seems daunting, but you can minimize the impact that your daily online habits have on that demand with little or no sacrifice. Here’s how:
Know what actions and settings affect data-center usage: Simply being aware that most of what you do online–using AI, streaming a video, playing a game—has an environmental cost will help you make better decisions about your online habits. You don’t have to give up what you find productive or otherwise of value; you just have to ask whether the action is important enough to add to the data-center workload.
Turn off AI features you don’t need: You might have active AI-powered features on your
device and apps that have no value to you. For example, if your phone tries to guess what
you want to type and you find it annoying, it might be an AI-powered predictive text feature that you can turn off. You can usually find what AI features are turned on for your device, browser or app in the settings menu. This Consumer Reports article explains how to to turn off AI features on a range of devices: bit.ly/3J1pEEg.
Practice good device hygiene: Periodically review the apps and data you keep on your
phone or store in the cloud and uninstall or delete what you no longer need. Data stored
online consumes data-center resources, and apps you don’t use could still be sending data to the cloud. This also ensures better privacy of your personal data.
Similarly, clean out old email messages. This might seem trivial, but all your email messages are likely stored in a data center. All the 5-, 10- and 20-year-old messages from the millions of U.S. citizens online represent a significant portion of the data stored online. The U.K. government thought this issue was important enough to issue a plea for its residents to delete old email messages to reduce data-center water consumption during a drought.
Turn on the option to process AI data only on your device: My Samsung phone offers this option, and I keep it on to reduce data sent to data centers and enhance my online privacy. Not all phones have this option.
Disable AI answers for your web searches: With Google, you can do this by adding -ai to
your queries. iPhone users will need to disable Apple Intelligence to eliminate AI search
results, and the above-mentioned Consumer Reports article explains how.