The Environmental Cost of AI

Looking for answers to questions about the universe? Need help writing code for a video game? Or just want to see a picture of kittens performing at a rock concert? Artificial intelligence (AI) tools can do all these things and more.

Are you looking for help? Maybe you need answers to questions about the universe. You might want to write code for a video game. Or you just want a picture of kittens performing a rock concert. Artificial intelligence (AI) tools can do these things and more.

 Electricity isn’t just needed to keep the batteries of phones, tablets, or laptops charged. “If you’re on a device at home, and you type in a query, that request is sent to a data center—a giant warehouse filled with computers,” says Jonathan Koomey. He’s been studying the environmental effects of computing for more than 30 years. Data centers run anything that requires the internet, like search engines, apps, video-streaming services, and gaming platforms. And that requires massive amounts of electricity.

We need electricity to charge the batteries of phones, tablets, and laptops. Many other activities require electricity too. “If you’re on a device at home, and you type in a query, that request is sent to a data center—a giant warehouse filled with computers,” says Jonathan Koomey. He has studied the environmental effects of computing for more than 30 years. Data centers run anything that uses the internet. That includes search engines, apps, video-streaming services, and gaming platforms. All of these require massive amounts of electricity.

There are more than 10,000 data centers worldwide—about half of which are in the United States. Currently, much of the electricity used to power them comes from fossil fuels, like oil, natural gas, and coal. Burning these fuels emits greenhouse gases that trap heat in Earth’s atmosphere, causing climate change. Tech companies like Amazon, Apple, Facebook, Google, and Microsoft have promised to produce zero greenhouse gas emissions by 2030 to help address this issue. To do so, they’ll need to transition from fossil fuels to renewable sources of energy, like solar and wind. Now experts worry that AI’s growing power demands will put those plans in jeopardy. In the past five years, for example, Google’s greenhouse gas emissions increased by nearly 50 percent. The company says that is largely because of increased energy use by data centers. Google says reducing future emissions could be challenging because of the rising energy demands of AI.

More than 10,000 data centers operate worldwide. About half of them are in the United States. Much of the electricity to power them comes from fossil fuels, like oil, natural gas, and coal. When these fuels burn, they emit greenhouse gases. The gases trap heat in Earth’s atmosphere and cause climate change. Tech companies are trying to address this problem. Amazon, Apple, Facebook, Google, and Microsoft have promised to produce zero greenhouse gas emissions by 2030. To do so, they’ll need to change from fossil fuels to renewable sources of energy. These include solar and wind. Now experts worry that AI’s growing power demands will threaten those plans. For example, Google’s greenhouse gas emissions increased by nearly 50 percent in the past five years. That’s mostly because data centers are using more energy, says Google. The company also says that reducing future emissions could be difficult. That’s due to AI’s rising energy demands.

“We don’t currently have enough renewable energy capacity to power all the AI tools,” says Luccioni. “And AI’s power requirements are growing faster than new solar plants and wind turbines can be built.” That’s why tech companies aren’t just investing in renewable sources to meet their future energy needs. They’re also turning to nuclear power—energy created from a reaction that splits apart atoms. Amazon and Google, for example, are backing the development of small nuclear plants that could be built quickly where needed.

While nuclear power doesn’t emit greenhouse gases, it’s not without risks. It produces toxic radioactive waste, which is difficult to dispose of. Accidents at nuclear plants can also expose people and the environment to dangerous radiation. These incidents are rare but do happen. One well-known example occurred at the Three Mile Island nuclear plant in Pennsylvania in 1979. A mechanical failure caused the release of radioactive material. There were no injuries or deaths, and the plant continued operating until 2019. Now the Three Mile Island facility plans to reopen after Microsoft made a deal to buy all the electricity generated there to power its AI data centers. Despite nuclear power’s safety concerns, the majority of scientists feel that its benefits, especially when it comes to reducing climate change, outweigh its risks, says Koomey.

“We don’t currently have enough renewable energy capacity to power all the AI tools,” says Luccioni. “And AI’s power requirements are growing faster than new solar plants and wind turbines can be built.” Tech companies are investing in renewable sources. But that’s not enough to meet their future energy needs. So they’re also turning to nuclear power. This energy is created from a reaction that splits atoms apart. For example, Amazon and Google are backing the development of small nuclear plants. These could be built quickly where needed.

Nuclear power doesn’t emit greenhouse gases. But it has other risks. It produces toxic radioactive waste, and this is difficult to dispose of. Accidents at nuclear plants are rare, but they do happen. An accident can expose people and the environment to dangerous radiation. One well-known example happened at the Three Mile Island nuclear plant in Pennsylvania in 1979. A mechanical failure caused the release of radioactive material. No one was injured or killed, and the plant continued to operate until 2019. Now Microsoft has made a deal to buy all of the plant’s electricity for its AI data centers. So the Three Mile Island plant plans to reopen. Even with the safety concerns, most scientists feel that the benefits of nuclear power outweigh its risks, says Koomey. That’s because it can help reduce climate change.

Some experts think the limited availability of space and resources needed to run data centers may force companies to develop AI that’s more energy efficient. That could potentially reduce the amount of greenhouse gases these tools produce. In fact, a new AI called DeepSeek created by a Chinese company has shown it’s possible for this technology to use far less energy than originally believed.

AI tools themselves could also help other industries reduce emissions. The technology could, for example, find ways to make construction, transportation, and agriculture more energy efficient. AI is also helping to locate new sources of geothermal energy—a type of clean, renewable energy produced by underground heat. And it can also help scientists monitor melting ice and deforestation and improve prediction of hurricanes and wildfires. “There are plenty of examples of real-world benefits from these tools,” says Koomey. But as to whether AI will be more helpful or harmful to the environment remains to be seen.

Luccioni says people—including you—should consider AI’s pros and cons when making decisions about its use. For instance, you don’t need to turn to a sophisticated AI tool, like ChatGPT, to help you solve a math problem. All you need is a pencil and paper or a calculator. “Just thinking about the environmental impacts of our actions, including using AI, is a good start,” says Luccioni.

Data centers need space and resources to run, and these are limited. So some experts think that companies will have to develop more energy-efficient AI. That could reduce the amount of greenhouse gases from these tools. In fact, a Chinese company has created a new AI called DeepSeek. It uses far less energy than people believed was possible for this technology.

AI tools could also help other industries reduce emissions. For example, AI could find ways to make construction, transportation, and agriculture more energy efficient. The technology is helping to find new sources of geothermal energy. This clean, renewable energy is produced by underground heat. AI can help scientists track melting ice and deforestation. And it can improve prediction of hurricanes and wildfires. “There are plenty of examples of real-world benefits from these tools,” says Koomey. But will AI be more helpful or harmful to the environment? That remains to be seen.

AI has pros and cons. You should consider both when making decisions about its use, says Luccioni. For example, are you trying to solve a math problem? You don’t need a sophisticated AI tool, like ChatGPT. All you need is a pencil and paper or a calculator. “Just thinking about the environmental impacts of our actions, including using AI, is a good start,” says Luccioni.