Various Fun Excerpts About Generative AI and the Environment
Air Pollution
“In South Memphis, Tennessee, a neighborhood called Boxtown is paying the price for the AI revolution with every breath. This historically Black community… faces a new threat: massive gas-powered generators installed by xAI — Elon Musk’s artificial intelligence company — that are poisoning their air. These generators power the “Colossus” supercomputer that trains Grok, Musk's competitor to ChatGPT. Environmental groups using thermal imaging discovered 33 gas turbines operating on site, emitting an estimated 1,200 to 2,000 tons of nitrogen oxides annually — compounds that cause respiratory problems and contribute to smog formation. This makes xAI one of the county's most significant pollution sources virtually overnight.”1
Water Usage
“As tech giants like Meta build data centers in the area, local wells have been damaged, the cost of municipal water has soared and the county’s water commission may face a shortage of the vital resource… If the local water authority cannot upgrade its facilities, residents could be forced to ration water.”2
“A data center like Meta’s, which was completed last year, typically guzzles around 500,000 gallons of water a day. New data centers built to train more powerful A.I. are set to be even thirstier, requiring millions of gallons of water a day, according to water permit applications reviewed by The New York Times.”2
“One data center company was asking for nine million gallons of water a day, equal to 30,000 households.”2
“The water used in data centers is often treated with chemicals to prevent corrosion and bacterial growth, rendering it unsuitable for human consumption or agricultural use. This means that not only are data centers consuming large quantities of drinking water, but they are also effectively removing it from the water cycle.”3
“One AI query is estimated to use 16 ounces of fresh water. ChatGPT estimates upwards of 10 million queries per day.”3
Energy Usage
“A single AI supercomputer draws enough power for 15,000 homes and requires continuous cooling and minimal transmission delays… The machines powering ChatGPT or Grok can't be hidden in distant countries; they need reliable power grids and fast internet connections to serve American consumers. But instead of distributing this infrastructure responsibly, Big Tech targets communities with the least power to resist.”1
“the process of training large language models (LLMs) consumes staggering amounts of electricity, contributing to increased carbon emissions.”4
“Large language models use about 30 times more energy than [websites] because it’s generating results to your prompts from scratch based on their training data, and that process takes a lot more computation.” 4