Opinion: What Are the Costs of AI on the Environment?

Editor’s note: This story contains the opinions of the author, not The Red Line Project.

AI disclosure: ChatGPT was used to generate research on this topic and create image prompts. Midjourney was used to create the images. The podcast was generated by NotebookLM.

In recent years, climate change has become one of the top issues the entire world is facing. Countries have pledged to lower carbon emissions, but this has had seemingly no impact as the environment is consistently being harmed by human activities. All the while, a new player has entered the game — artificial intelligence.

It is often overlooked because of its recent development and a lack of standardized measurement systems, but make no mistake, AI is a large carbon emitter and will only get worse as it advances.

Not only does AI emit carbon dioxide, but it also requires extreme energy use, high levels of water intake, and produces extensive waste that harms the environment. And AI has only just begun. With its current progress, by 2050 the largest carbon emitter won’t be big oil or coal companies, but the small black dot sitting on your dresser that goes by the name Alexa.

Currently, AI’s carbon footprint is hard to track. Besides the lack of measurement of its emissions, big tech companies often do not publicly share the carbon footprint of training and using their models. According to the Anatomy of an AI System, the world’s biggest cloud computer company, Amazon Web Services, is completely non-transparent about the energy and carbon footprint of its massive operations.

However it goes unnoticed by the general public, AI does produce a significant amount of carbon emissions. A study done by the University of Massachusetts Amherst in 2019 showed the estimated carbon footprint of training a single big language model is equal to around 626,000 pounds of carbon dioxide emissions. Those emissions are equivalent to flying 300 round-trip flights from San Francisco to New York City. It’s amount is roughly five times the lifetime emissions of the average American car.

PODCAST: Listen to this AI-generated podcast about the effects of AI on the environment covered in this article.

As found by OpenAI researchers, since 2012 the amount of computing power required to train new models has doubled every 3.4 months, where it consumes thousands of megawatt hours of electricity per training process. By 2026, data center energy consumption is set to take 6% of the nation’s total electricity usage, according to the Harvard Business Review.

“A generative AI query uses up to 15 times the amount of energy as a traditional web search,” said USC Professor Kate Crawford at a UIC AI Symposium in October. “Image generation and video generation use much more energy again.”

The more energy needed; the more carbon emissions will increase. It is expected that emissions from the Information and Communications Technology industry as a whole will reach 14% of global emissions by 2040, primarily from data centers.

Beyond carbon emissions and energy use, AI has an extreme water thirst. Data centers need enormous amounts of freshwater to cool the servers 24/7.

“A single large data center can consume up to 5 million gallons of water a day,” Crawford said.

In her lecture, Crawford also discussed Google’s data center in Oregon. A lawsuit from locals in the area revealed that the center was using one-third of the district’s freshwater resources for its own cooling needs.

This large water intake comes at a time where regions enter freshwater scarcity, like in parts of Latin America and in southern states in the US, where climate change has significantly decreased the amount of water in rivers and reservoirs and where droughts occur more frequently.

As AI grows and data centers increase, so too will the amount of water it requires, further depleting communities that are already suffering.

Artificial intelligence also directly harms the environment through its waste. E-waste produced by AI technology includes servers, GPUs and computing chips, all of which have short life cycles, and are therefore constantly being replaced by new technologies. This results in a high level of hazardous materials being thrown into the landfills.

This can include lead, mercury and cadmium, which can contaminate the air, water and soil. Data in Statista show that in 2022, 62 million metric tons of e-waste was generated across the world, while waste management remained inadequate. However, the waste of AI technology is not the only issue, so is the mining of the materials for that tech.

Through mineral acquisition, many unwanted elements are also retrieved and disposed of above land, though they would normally remain underground and away from humans and animals.

According to the Anatomy of an AI System, the mining of dysprosium and terbium in China, both used in multiple tech devices, produces more waste than valuable minerals. Only 0.2 percent of the mined clay contains the elements, meaning 99.8 percent of the Earth removed in the mining is discarded as waste called tailings. The tailings are dumped back into hills and streams, adding more pollutants to the environment, like ammonium.

Despite all of the challenges AI poses on the environment, we know its integration into our world has only just begun. The good news is that we can move forward with AI sustainably.

The best way to do so is to transition to renewable energy and end the use of fossil fuels. By powering data centers with sources such as solar, wind or hydroelectric, tech companies can dramatically reduce AI’s carbon footprint. Due to the Tech Workers Coalition that marched in demand for tech giants such as Microsoft and Google to reduce emissions by 2030, companies are already on the path to using 100% renewable energy sources.

Additionally, instead of discarding all e-waste, companies should be proactively reusing as much hardware as possible and properly recycle all unwanted materials.

The path forward is not easy. AI technology takes seemingly more than it gives, and despite all of its wonders, it needs to become more sustainable before we damage our Earth beyond repair.

That will take time and also trust in companies and their commitment toward renewable energy. However, we must not only rely on large companies, but also researchers, policymakers, and environmental organizations to work together and come up with the best solutions. By doing so, we can reap the benefits of AI while also protecting the environment and move into a more sustainable future.