Sam Altman Responds to AI Water Consumption Concerns
OpenAI’s CEO, Sam Altman, has downplayed fears surrounding artificial intelligence’s water usage, calling such worries “bogus.” He emphasized that “humans also consume energy,” amid claims suggesting that OpenAI’s ChatGPT uses a gallon of water for simple responses.
During the India AI Impact Summit, Altman stated these assertions about water consumption were “completely false and completely outrageous.” He asserted that they lack any factual basis.
He also addressed comments made by Microsoft founder Bill Gates, who had mentioned that, eventually, AI could improve its energy efficiency due to the efficiency of the human brain. Altman argued, however, that comparing the energy needed to train AI models with human development is misleading. “Training humans also requires a lot of energy,” he remarked, pointing out that gaining intelligence takes around 20 years and all the food one has consumed.
He suggested that a more meaningful comparison would be evaluating how much energy is used when ChatGPT answers questions compared to human responses. By that measure, he believes AI might be approaching similar energy efficiency.
While he did dismiss the specific water usage claims, he admitted that worries about overall energy consumption are valid. Altman pointed out that as AI’s prevalence grows, the demand for sustainable energy sources like nuclear, wind, and solar power becomes increasingly urgent.
A study indicated that by 2023, data center energy consumption had already reached levels comparable to the total power usage of Germany or France.
Following Altman’s remarks, some online reactions criticized his comparison of humans and AI. For instance, Sridhar Vembu, co-founder of Zoho Corporation, emphasized that equating technology with humans isn’t acceptable.
Investments in artificial intelligence are booming, both from tech companies and governments. Recently, President Trump announced a “Technology Corps” aimed at promoting American AI globally and training engineers for the field.
Last year, the Trump administration introduced a plan dubbed Stargate, intending to invest $100 billion in extensive data center projects across the U.S., though progress has reportedly been slow.
Environmental organizations are voicing concerns over fast-tracking approvals for data center constructions, while local communities are protesting new facilities, fearing they will overburden power grids, increase costs, and stress nearby water resources.
A proposed $1.5 billion data center in San Marcos, Texas, was recently rejected by the City Council after significant public dissent.
Data centers, known for their high energy demands, use large amounts of water for cooling, with estimates suggesting that responding to a 100-word AI prompt requires about a bottle of water. Medium-sized data centers can consume around 300,000 gallons of water daily, equivalent to what 1,000 American households might use.
Some newer data centers are designed to minimize water usage entirely. Nonetheless, projections indicate that water consumption for cooling in data centers could more than triple over the next 25 years due to rising computing demands. This was highlighted in a recent report from Xylem and Global Water Intelligence.
