Surge pricing: Decoding the rising cost of our digital lives

That’s how many emails we’re sending every day, in 2025, according to a recent study by technology market research company Radicati. If that seems high, just scroll through the junk mail and newsletters in the rarely accessed folders of your inbox.

Every email has a complex carbon footprint. There is the energy use on the client device: the personal phone, PC or tablet. The energy required for transmission: via ethernet, wifi or mobile networks. And there’s everything that happens at the data centres, where servers process each message. This involves storage, checking for spam, relays from the sender’s server to that of the recipient.

Energy is consumed when the message is read too. Add to that the energy it takes to sort, download attachments and run background security programs.

It is a sequence of events we now take for granted.

It all adds up.

In 2010, researcher and carbon-accounting pioneer Mike Berners-Lee — coincidentally the brother of Tim Berners-Lee, inventor of the World Wide Web — wrote a book titled How Bad are Bananas? The Carbon Footprint of Everything.

In it, he estimated that each email has a footprint of at least 0.03 gm of CO2eq.

At 376.4 billion emails a day, that’s an annual toll of 4.12 million metric tonnes of CO2eq. Or, the equivalent of flying from Mumbai to New York and back more than 700,000 times.

***

Email’s energy footprint is massive, but is dwarfed by the footprint of online searches.

Google’s sustainability reports indicate that a single search uses 0.0003 kWH of energy, which would translate to at least 0.14 gm of CO2eq. That’s almost five times the footprint of a text-only email. And Google processes more than 5 trillion searches a year.

Computing has always been as much about energy as processing power.

Early IBM mainframes were installed in dedicated rooms, cooled by large fans as well as water. By the 1960s, Seymour Cray was designing his iconic supercomputers with two basic principles. The first was: Remove heat. (The second: Ensure all signals arrive on time.)

The old mainframes were remarkably inefficient, requiring hundreds of kWH of energy for simple operations.

The 1990s transformed computing. The invention of the World Wide Web laid the foundation of today’s tech world. 16-bit architectures evolved into 32 bits and finally into 64 bits, dramatically increasing the amount of memory accessible to the computer’s CPU, paving the way for computation-intensive operations such as video processing, advanced gaming and photo editing; things that we now take for granted.

The web popularised the client-server paradigm, paving the way for server farms, cloud computing and data centres; all of which are power-hungry. As computing efficiency has improved, the complexity of the operations we perform on our devices has more than kept pace.

Behind every Reel, song, streaming platform and YouTube video, as well as every tweet, chat message and hour of gaming, are server farms and data centres humming without pause, drawing hundreds of megawatts of power, to run tens of thousands of processor cores (and to run cooling systems to keep those cores from melting).

***

The largest data centre in the world, a China Telecom facility in Hohhot, is spread across more than 10 million sq ft, and consumes 150 MW of power — as much as a small town of 25,000 people — and has the annual CO2eq emissions of 250,000 cars.

Meanwhile, the live-streaming and online shopping, videoconferencing, Slack and Teams meets are just the consumer segments of this virtual world.

There are also the business users: banks and corporations, defence firms, governments, healthcare chains, telecom networks. Since the details of their usage is not often publicly available, it is hard to estimate the impact that the cloud as a whole has on the planet.

***

The Aughts saw the arrival of another energy-intensive product: cryptocurrency.

At its core, crypto-mining is a competitive process in which powerful computers race to solve a complex cryptographic puzzle. The winner gets to update the shared transaction ledger (the blockchain) and is rewarded with new cryptocurrency. Puzzles are regularly updated to become increasingly difficult, requiring more and more computational power to solve.

The US Department of Energy estimated that crypto-mining consumed 50 terawatt hours in the US alone, in 2024. While the US is the biggest player in this field, Russia, China, Kazakhstan and Canada also have significant cryptocurrency mining operations.

Now there’s artificial intelligence as well.

OpenAI launched ChatGPT on November 30, 2022. In five days, it had hit a million users. By January 2023, over 100 million people were using it, making it the fastest-growing software application in history. Other AI interfaces followed: Google’s Gemini, Facebook’s Meta AI, xAI’s Grok, Perplexity, DeepSeek and others.

In January this year, OpenAI CEO Sam Altman blogged about energy use. “People are often curious about how much energy a ChatGPT query uses; the average query uses about 0.34 watt-hours, about what an oven would use in a little over one second, or a high-efficiency lightbulb would use in a couple of minutes,” he wrote. It would also use “roughly one fifteenth of a teaspoon” of water, he added.

This accounts only for the energy required to generate ChatGPT’s responses.

In addition, there is the infrastructure required for training: datasets made up of billions of words and trillions of tokens or packets of information, chips designed to handle real-time analysis and reanalysis of information from across this dataset; all of which translates to more servers, cooling and networking. All of which was neatly sidestepped in Altman’s post.

Even excluding the training costs, Altman’s figures translate to about 0.153 gm of CO2eq per query, more than 500 times the cost of a Google search.

Studies, meanwhile, indicate that the energy costs for a medium-length text response by GPT 5 could be as high as 18 watt-hours.

Computing is energy-intensive. And that cost is climbing.

***

It says something about the way our world is organised that the tech sector still doesn’t even make it to the list of Top 5 most energy-intensive industries.

Transportation, manufacturing, construction, agriculture, the manufacture of chemicals, paper and pulp, and food processing, all rank higher on the emissions scale. Unlike some of these, the costs for the computing world are seen as so high that several tech giants are working to reduce this footprint.

As for us, perhaps the best thing users can do is follow good computing hygiene. Unsubscribe where you can. Download, don’t stream. Stop scrolling.

And maybe go out and smell the roses every now and then, instead of staring at a screen all day.

(K Narayanan writes on films, videogames, books and occasionally technology)