Surprising AI Environmental Impact vs Regular Tech Use

What to Know

The negative environmental impact of AI includes wasting water and energy, carbon emissions, and e-waste.
Sending emails, streaming content, online banking/shopping, and using social media also harms the environment.
While tech can damage the environment, it can also be used to protect it.

Society has been abuzz with horror stories about the AI impact on our environment, which is valid. However, almost all technology use comes with an invisible carbon footprint, and some of the regular ways we use technology and spend time online are even more harmful than using generative AI.

How Is the Human Impact on the Environment Measured When It Comes to AI?

Before I dive into the research that's been done on this, I want to point out an important issue with environmental impact studies. It is really hard to measure AI water usage, energy usage, and carbon footprint. So the answer to, how much energy does AI use, gets complicated because not only are there lots of moving parts, but also a lack of clear regulations. On February 1st, 2024, Massachusetts Senator Edward Markey helped introduce a bill that would "require the federal government to assess AI’s current environmental footprint and develop a standardized system for reporting future impacts." A similar bill was passed in the EU, called the European Union’s “AI Act.” It requires companies to accurately report how much energy they consume and what resources they consume, plus any impact that their systems have on the planet. Unfortunately, the bill did not pass, but many politicians are currently lobbying for more environmental regulations regarding AI.

Without such rules in place, we can only speculate what the impacts of AI are. This is because using AI involves physical data centers that require huge amounts of power, but also consume large quantities of water for cooling purposes. Then there's all the energy and resources spent when the public uses these models to ask questions or generate content. A slightly more hidden aspect is all of the training that AI models require, which takes months and is a significant power drain. So when the Guardian looked into the greenhouse gas emissions that companies like Google, Microsoft, Meta, and Apple reported between 2020 and 2022, they were as much as 662% higher than reported.

Finally, it is important to note that there are eco-friendly ways to power data centers. When we consider the amount of CO₂ emissions produced for every kilowatt-hour (kWh) used, it is important to keep in mind that various countries, states, and regions use different resources for their energy. Some places mostly use coal or gas, while others might have more renewable energy sources like solar or wind. This is why it is so important for more and more energy to come from renewable sources, producing less CO₂ emissions despite using these data centers.

How Much Water Does AI Use Compared to Other Online Activities?

Water comes into play because data centers use a lot of it to cool down their infrastructure, and while water is an important resource, many studies focus on CO₂ emissions rather than the overall AI water consumption. Fortunately, it is relatively easy to calculate how much water is used based on the kWh consumed. One important factor is that some data centers recycle water so that a single gallon of water can be used for multiple rounds of cooling. However, no matter how it is 'recycled,' there will always be some water loss. Eventually more water will be needed to keep things running. The data centers that belong to Oracle AI, which powers ChatGPT, happen to use a closed-loop cooling method. According to the Environmental and Energy Study Institute, "large data centers can consume up to 5 million gallons per day, equivalent to the water use of a town populated by 10,000 to 50,000 people." The cool thing about AI is that it can help humans develop better ways to recycle resources to make data centers more efficient and eco-friendly so that we can keep using them without further harming the environment.

While it's becoming more commonly known that AI has a huge environmental footprint, so do many other activities that many of us practice every day. For example, a study from November 2019 revealed that "the UK could reduce its carbon output by over 16,433 tonnes simply by each adult sending one fewer ‘thank you’ email a day." Since this study is from before COVID and the AI gold rush, it is safe to assume that the number of 'unnecessary' emails sent in the UK in 2026 is a lot more than the 64 million daily emails sent in 2019 that simply had one or two word responses including "thanks," "received," "cheers," and "lol".

The International Energy Agency calculated that back in 2022, AI (in addition to data centers and cryptocurrency) used around 460TWh (terawatt hours) of electricity. Their prediction for 2026 is that these same three systems will use over 1,000TWh. But data centers aren't just for AI, they also store, host, and transmit content required for streaming. Forbes reported in December of 2025 that, according to the TRG Datacenters, "watching Netflix for an hour creates 500 times more CO₂ than sending two text prompts to Gemini or ChatGPT."

Social media is also a big drain on resources. Compare the Market has a great tool that you can use to calculate your own social media carbon footprint based on how many minutes you spend on each platform every day. Not too surprisingly, apps like TikTok And Instagram generate a large carbon footprint because of their focus on short video streaming.

According to a study quoted in an article by Me Mother Earth, TikTok has a carbon footprint of approximately 57,597gCO₂Eq per year, which is the same amount of CO₂ emitted during a 143 mile-drive in an average gas-powered sedan. The second biggest CO₂ emitter from Me Mother Earth's list of platforms was Reddit, using around 54,312gCO₂Eq per year (compared to 135 miles). Going down the list we have Pinterest (68 miles), Instagram (57 miles), Snapchat (47 miles), Facebook (43 miles), and Twitter (33 miles).

Pro Tip:

CO₂Eq stands for carbon dioxide equivalent and converts the impact of greenhouse gases such as methane, nitrous oxide, and nitric acid into a carbon dioxide equivalent.

Unfortunately, it isn't just entertainment that's power-hungry… video conferences via Zoom and watching tutorials on YouTube are costly to our planet, too. According to the TRG Datacenters's study, streaming on Netflix or YouTube in HD for just one hour consumes around 0.12 kWh which results in 42 g CO₂ emissions. To put things into perspective, a car emits this much driving just 0.1 miles (around 160 meters). So while watching Netflix isn't as bad as driving your average gas sedan, it is a lot compared to other online activities.

Pro Tip:

Terawatt-hours are used when measuring energy globally, while kilowatt-hours are used when we discuss the consumption of energy within a household. For reference, 1 TWh is equivalent to 1,000,000,000 kWh.

So while generative AI gets a bad rap, the average text-to-video generation to create a 6- to 10-second video will use around 0.05 kWh, which results in around 17.5 g CO₂ of emissions. While this is less than watching half an hour of Netflix, it is just a bit more than being on a Zoom call for one hour. Compared to this, sending a short email without an attachment uses around 0.0133 kWh, which creates 4.7 g CO₂ emissions. This means that generating one AI image is 'greener' than sending a short email, because the average image generation uses around 0.003 kWh, equalling 1 g CO₂ emissions. Activities such as making a voice assistant query via Siri, doing a Google search, and giving an AI chatbot a prompt each use under 0.0005 kWh, which produces 0.175 g CO₂ emissions. If you use Gemini, you'll be happy to hear that two Gemini prompts 'cost' us less than one Google search, because they use around 0.00024 kWh, which equals 0.084 g CO₂ emissions.

Surprisingly enough, some of the things we do online are more environmentally-friendly than we might imagine. Online shopping, for example, seems like a huge use of energy, but traditional retail can be much worse. The New York Times points out that having "one van delivering 50 packages is much more efficient than 50 people driving to the store." That being said, doing a Zoom call rather than driving to work/school/bank/doctor's office might result in fewer CO₂ emissions. Again, it depends, because some places have lots of public transportation, or a higher proportion of electric vehicles or hybrid cars. Likewise, certain stores only source their products locally, while others ship them over by plane from overseas. All these little things factor into the environmental footprint.

To wrap things up, an individual like you or me wastes a negligible amount of water per day when using AI. We waste more water on social media or streaming; however, even an hour on TikTok uses up to 16.91 US fluid ounces (half a liter) of water. While AI and social media are still a lot less harmful for the environment than driving, the numbers do add up. According to a study by Review, "the average American spends 3 hours 49 minutes per day using streaming services." Another study from May 2024, estimates that 60% of the global population has internet access and the average user spends "over 40% of their waking life on the Internet." Considering that there are currently over 8.3 billion people on Earth, these numbers really put things into perspective. To stay in the loop on the latest Apple tech, don't forget to check out for our free Tip of the Day newsletter.

Ways to Reduce Carbon Footprint While Still Using AI & Being Online

We are often told to spend less time online, use AI less, stream less, etc. But that isn't always applicable in today's society. As individuals, it is hard to make a significant impact on our carbon footprint outside of using our tech more mindfully. On the other hand, we as a society can do a lot to improve the situation.

If governments and companies create stricter rules about using more renewable energy sources, we could see a real difference in our overall carbon dioxide emissions. We are already seeing AI data centers use solar energy and recycle water. Some companies are also more environmentally friendly or even carbon neutral, which means that they offset their CO₂ emissions by planting trees, etc. Google actually became carbon neutral in 2007, but with the rise of AI, it has stopped claiming this status as of 2023. They are currently hoping to become carbon neutral once again by 2030 alongside Apple and many other tech giants.

Staying up to date on technological and environmental news can help us be more informed, which helps us make better choices when voting in local elections. There are also petitions that can be signed, local representatives that can be contacted regarding certain issues, and many ways to spread awareness so that more people can make better choices and take actions towards a greener future.

If you want to explore your personal carbon footprint or learn more about how much carbon is produced or water is wasted with certain activities, there are lots of websites that contain calculators based on various studies. Forterra gives you very specific prompts and options to figure out your overall carbon footprint as an individual.

Now that you have a better understanding about AI's environmental impact, you can make informed decisions going forward on how to reduce your carbon footprint. I was honestly surprised that many of my daily activities are hurting the environment way more than using AI to generate an image or answer a burning question. Of course, if you use AI for everything all day long, your carbon footprint might be bigger than that of someone who casually scrolls on Facebook and watches a movie on Netflix.

FAQ

Is ChatGPT bad for the environment? Yes. But so are most things we do online, like streaming, video calls, and using social media.
How is ChatGPT bad for the environment? Generative AI takes a lot of power and water to train and use.
Why does AI use water? Water is used to cool down the powerful data centers that keep AI running smoothly.
How much water does AI use per day? According to EESI (Environmental and Energy Study Institute) large AI data centers use as much water every day as a town populated by 10,000-50,000 people, which is as much as 5 million gallons every day.

Top image credit: Ravlyyy / Shutterstock.com