Data centers, which serve as the vital foundation for artificial intelligence (AI), are increasing their energy demands every year. According to experts’ forecasts, by 2030 they will consume a staggering 945 TWh of electricity annually. For comparison: this is more than seven times Ukraine’s current annual energy consumption.
The cost of progress, with the rapid development of AI, could deal a “severe” blow to Earth’s fragile ecosystem. Illustrative photo: Unsplash
However, energy consumption is just the tip of the iceberg. Every request processed leaves behind not only a carbon footprint but also a “water footprint” due to server cooling, as well as a “land footprint” from infrastructure construction and supply chain maintenance.
Water and Earth
A new study by the United Nations University (UNU) highlights a critical issue: the environmental impact of AI is being measured incorrectly. Researchers typically focus solely on greenhouse gas emissions, while ignoring the depletion of other resources. According to experts’ estimates, by the end of the decade, AI will require as much water as is needed to meet the basic domestic needs of 1.3 billion people.
AI is depleting the planet’s water resources. Photo: Unsplash
In terms of land area, the infrastructure will expand to over 14,500 square kilometers—twice the size of the Jakarta metropolitan area. The report’s authors warn that even the transition to “green” energy to reduce emissions could backfire, as some renewable sources require even more land and water resources.
The environment’s main enemy
There is a common misconception that the most resources are consumed during the training phase of new AI models. However, statistics show that 80% to 90% of energy is actually consumed during the daily use of these services. Just one popular platform processes approximately 2.5 billion requests per day, consuming hundreds of GWh per year.
Data center. Illustrative photo: Unsplash
The scale of the computational cost depends directly on the complexity of the task. Text classification is considered the most “energy-efficient,” but as soon as it comes to generating a single image, energy consumption increases a thousandfold. Creating videos requires even more massive computational power. Even optimizing algorithms doesn’t save the day: making the process cheaper only encourages people to use neural networks more often, creating the so-called “rebound effect.”
Environmental inequality and piles of waste
The whole world benefits from the advancements in AI, but the environmental consequences are clearly concentrated in certain regions. In many countries, new data centers are placing a critical strain on local power grids and causing water shortages, even during droughts.
Another crisis is coming: an avalanche of electronic waste. By 2030, the AI industry will generate approximately 2.5 million tons of end-of-life equipment annually. This toxic burden will fall primarily on low-income countries, which lack facilities for the safe recycling of electronics. The mining of rare minerals for new processors also destroys ecosystems and exacerbates social inequality in mining regions.
Digital divide between countries
The burning of electronic waste in Africa is a massive environmental and humanitarian problem, best known for the Agbogbloshie landfill in Ghana. For years, millions of tons of electronic waste from around the world have been illegally dumped there. Photo: npr.org
The technological boom is exacerbating global inequality. According to the report, more than 90% of specialized AI computing power is monopolized by just two countries—the United States and China. In contrast, more than 150 countries worldwide lack their own specialized infrastructure. This violates the basic principles of environmental justice: many nations are forced to endure environmental damage but do not reap the economic benefits of AI development.
On the path to ecological balance
UNU researchers emphasize that their goal is not to halt progress, but to make it safe. The scientists are calling for the immediate establishment of an “ecosystem for responsible AI” based on transparency, fairness, and sustainable use of resources.
Programmers, for their part, should also use AI more sparingly. Photo: Unsplash
Governments are advised to make data center planning part of their overall water and land management strategy. Developers should create software that is more efficient and user-friendly. Ordinary users can also help by choosing tools that are less harmful to the environment. The future of the digital age will depend not only on the ingenuity of algorithms, but also on the conscious decisions we make today. Because we are already responsible for the state in which we will leave the Earth to our descendants.
We previously reported on how artificial intelligence learned to measure the expansion of the universe.
According to un.org