AI has taken the world by storm, infiltrating nearly every industry from medical care to agriculture. However, the energy necessary to sustain these technologies translates to excessive water consumption.
Amid climate change, resource conservation is of the utmost importance. With nearly 2 billion people experiencing water insecurity worldwide, excessive consumption is the bane of progress. Now, the question focuses on how AI programs can continue operations without furthering the crisis.
According to data, training one AI program uses around 1,300 megawatt hours (MWh) of electricity, or the sum of power consumed by 130 homes annually in the US. Moreover, the computing power for AI is estimated to double every 100 days for the next five years. Generating such electricity releases heat, and many data centers utilize water evaporation to lower temperatures.
The US has yet to set strict AI utilization and employment limitations. Furthermore, the law does not require AI companies to disclose their environmental impact, unlike many other large organizations across the globe. This lack of oversight allows dishonest developers to manufacture incorrect water and energy consumption reports.
Smaller AI models can perform simple tasks like summarizing tasks. However, most people use AI for more intensive undertakings, such as writing essays, analyzing data, or producing audio. The larger devices required for these requests consume significantly more resources.
In response to growing concerns, some AI data centers have adopted sustainable practices to reduce consumption or recycle water.
For example, one Illinois facility implemented mixed efforts to reduce the scaling potential of its cooling system influent. Consequently, water consumption dropped from 25-30% to 12-15%, saving roughly 12MM gallons of water.
Sustainable technologies like the Thermosyphon Cooler Hybrid System are also emerging, making for an easier solution to high water use. Findings suggest this “chiller-less design, component-level warm-water liquid cooling, and water heat capture and reuse” approach can save up to 2.10 million gallons in 2 years.
