If a student asks ChatGPT 10-50 questions to help them with their homework, artificial intelligence (AI) will consume approximately 500 milliliters of water, which is the volume of an average water bottle.
The Conservation Foundation affirms this fact and reports that AI technologies’ water consumption is a concern. The use of AI generates a significant amount of heat, which is then released into AI data centers. Cooling down the data centers requires an enormous amount of water.
The AI water supply
Large data centers can consume up to 5 million gallons of water per day, according to researchers at the Lawrence Berkeley National Laboratory. If current trends continue, the global demand for water generated by AI is expected to reach 6.6 billion cubic meters by 2027.
An AI data center is a specialized infrastructure designed to support intensive AI and machine-learning workloads. It utilizes high-performance servers equipped with AI accelerators, enabling advanced networking and high-speed communication.
“It doesn’t have to be good quality water to cool the AI data centers,” said Thanos Panagopoulos, associate professor of computer science at Fresno State. “Any type of used water will work for the cooling.”
Panagopoulos said that, if anything, AI is a very small portion of the total water usage in the industry compared to agriculture, which is at the top.
He said AI can offset water usage with precision technology-based irrigation, which helps avoid irrigating too much or too little. AI can help irrigate exactly what is needed, and by doing so, it can help save a lot of water.
The energy matter
President Donald Trump’s AI plan aims to accelerate the construction of AI data centers, bypassing public protections, environmental laws and critical oversight, as reported by Truthout.
A data center campus is a site with multiple buildings to house operating data centers. A data center campus, using one gigawatt of electric power annually, would use more power in a year than consumers use in Alaska, Rhode Island or Vermont.
Despite Trump cutting $4 billion in federal funding for California’s high-speed rail system, private investors are negotiating with top high-speed rail executives the possibility of connecting Fresno’s rail system to a solar grid and establishing AI data centers, according to Fresnoland.
High-speed rail executives are seeking public-private partnerships to enhance the viability of the $120 billion train project. AI data centers could be part of the solution.
“I think it is entirely possible that an AI data center could be built in Fresno,” said Naomi Bick, assistant professor in the Department of Political Science at Fresno State.
Bick said that AI data centers consume a significant amount of energy, and more energy would need to be generated or purchased to account for the increased demand. Even if the solar grid supplements the electricity, there is still an energy cost and use to be provided for.
She said that, just like the energy, more water would have to be considered for use.
Fresno’s water consumption is about 108 million gallons of water per day, based on 2020 data.
Finding additional water support for data centers in the San Joaquin Valley could be challenging. Nearby farmers and rural communities struggle to replenish groundwater aquifers under the state’s groundwater law, which has been in effect since 2014, according to Fresnoland.
California’s Silicon Valley is one of the world’s largest and oldest data center markets. Still, analysts have widely stated that the state’s high land costs and expensive electricity will limit its ability to capitalize on the current wave of demand for AI data centers, according to Reuters.
As a solution, several of the proposed new projects aim to establish AI data centers in Fresno and Contra Costa County instead.
However, communities across the U.S. are pushing back against AI data centers, arguing that they consume more resources than the benefits they offer.
Racial justice concerns
Truthout reported that Elon Musk’s $12 billion AI company, xAI, is emitting toxins in Black communities in Memphis, Tennessee.
Musk’s supercomputer, Colossus, is situated a few miles from Boxtown, a predominantly Black community and a town with a population of approximately 3,000. Nearly half of the residents have an annual household income of less than $25,000, and the cancer rates are four times the national average.
According to the Southern Environmental Law Center, the AI data center in Memphis has fueled 35 unpermitted gas turbines that produce smog-forming pollution and harmful chemicals, including formaldehyde. Nitrogen dioxide levels have increased by 79%. Nitrogen dioxide is linked to respiratory illnesses. Memphis has more emergency visits than other cities in the state.
“We know environmental racism and environmental justice issues are bound in the South and here in the Central Valley,” Bick said.
She stated that Fresno has a history of exclusionary zoning, where people of color and low-income families live in specific parts of town, located where heavy industrial zoning tends to be. They end up suffering from more pollution effects, so the building of an AI data center is a continuation of a long series.
“It’s not that they are necessarily choosing places of lower income, but that’s the sort of places that these types of zoning are allowed to happen,” Bick said.
Bick said that there tends to be less political resistance in those regions because the demographic is more working-class and tends to be less politically organized, and therefore less likely to push back against such measures. In contrast, people in more affluent areas are more likely to do so.
“It’s more of a political issue than a technical one,” Panagopoulos said. “AI, like any tool, can be used in many ways, and it’s up to us to decide how to use it.”
