AI Data Centers and Energy Consumption
Generative AI technology is causing a significant increase in energy consumption worldwide. This type of AI requires large data centers to store and process vast amounts of information. Currently, there are more than 6,000 data centers across the globe, and this number is expected to rise to almost 8,400 by 2030. These data centers consume a lot of electricity, which leads to higher carbon emissions. According to reports, a single query on ChatGPT consumes almost 10 times more electricity than a basic internet search. This excessive energy consumption makes AI systems one of the fastest-growing contributors to carbon emissions in the tech industry.
Data centers use both electricity and water for cooling purposes. In 2022, Microsoft used around 52 million liters of water to train its ChatGPT model in Iowa, USA. This amount represented 6% of the entire city’s water usage for that month. Similarly, Google’s carbon emissions increased by 48% between 2019 and 2024 due to higher data center energy consumption. Water usage by Google’s data centers also rose by nearly 88% during the same period. The increased demand for AI services like image generation, language translation, and speech recognition further amplifies the need for more powerful hardware, leading to even higher energy usage.
The location of data centers also plays a crucial role in their environmental impact. Data centers built in regions with limited renewable energy sources depend more on fossil fuels, making their carbon footprint even larger. In countries where coal or gas is the primary source of electricity, the environmental harm caused by data centers can be significantly higher.
AI-Driven Water Consumption and Environmental Strain
Generative AI models require water for cooling equipment in data centers. This water is taken from local water supplies, putting pressure on communities. Some regions, like Arizona, Spain, Taiwan, and Singapore, have limited freshwater resources. Data centers in these areas add extra strain on already scarce water supplies. The increasing demand for water by AI infrastructure raises concerns about water scarcity and its effects on both ecosystems and human populations.
Studies show that writing a short email of 100 to 250 words using a large AI model can use between half a liter and three liters of water, depending on the location and weather conditions. Many companies only report direct water usage for cooling, but indirect water consumption during electricity generation also adds to the environmental impact. This hidden water consumption makes it difficult to assess the full environmental footprint of AI technologies.
In Ireland, data centers already consume 21% of the country’s electricity, more than the combined energy usage of all urban households. The growing demand for AI technologies is likely to worsen this situation, affecting both the environment and local communities. The impact on water resources can also disrupt local agriculture and biodiversity, especially in areas facing frequent droughts.
Data center operators are exploring various cooling technologies, such as air cooling and liquid immersion cooling, to reduce water consumption. However, these alternatives often come with higher energy demands or technological limitations, making widespread adoption difficult. Transparency in reporting both direct and indirect water usage is essential to understanding the full scale of the environmental burden caused by AI.
Carbon Emissions and Public Health
Data centers not only consume energy but also produce significant amounts of carbon emissions. From now until 2030, data centers are expected to release around 2.5 billion tons of carbon dioxide into the atmosphere. This is roughly 40% of the annual carbon emissions of the United States. The emissions from AI systems contribute to the rising levels of greenhouse gases, accelerating climate change.
A study found that training one large AI model like Meta’s Llama 3.1 could produce air pollution equivalent to over 10,000 car trips between Los Angeles and New York City. Such emissions contribute to global warming and negatively impact public health. The cost of air pollution caused by data centers in the U.S. is estimated to be around USD 20 billion per year. These emissions also increase the risk of respiratory diseases, heart problems, and other health issues among nearby populations.
While many tech companies claim to aim for net zero emissions, they often rely on purchasing carbon credits instead of reducing their own emissions. Carbon credits allow companies to offset their emissions by investing in environmental projects elsewhere, but this does not reduce pollution at the source. As AI technology continues to grow, the environmental and social costs are becoming more visible worldwide. Without stricter regulations and genuine efforts to cut emissions at the source, the expansion of AI systems could significantly hinder global efforts to combat climate change.
Public awareness and pressure on tech companies to adopt more sustainable practices are increasing. Some companies are exploring renewable energy sources and energy-efficient hardware to reduce their carbon footprint. However, the rapid pace of AI development often outpaces these sustainability measures, making it difficult to balance technological progress with environmental responsibility.
