Sam Altman Rejects AI Water Usage Claims While Comparing Model Training Energy to Human Development

Sam Altman labels AI water usage claims as crazy and compares model training to human life. Explore the energy debate shaping the 2026 tech landscape.

By: AXL Media

Published: Feb 24, 2026, 5:22 AM EST

Source: The information in this article was sourced from

The Dispute Over Resource Intensive Computing

The global debate surrounding the environmental footprint of artificial intelligence has intensified following provocative statements by Sam Altman at an industry summit in India. The OpenAI executive countered prevalent narratives that portray data centers as massive consumers of water resources, labeling such claims as crazy and unfounded. While critics point to the vast amounts of liquid cooling required to prevent server overheating, Altman maintained that the current portrayal of water usage is a complete distortion of the actual operational requirements of high-performance computing facilities.

Conflicting Reports on Future Water Demand

Altman's dismissive stance arrives despite recent industrial findings that suggest a looming crisis in cooling infrastructure. A report released last month by Xylem and Global Water Intelligence projected that the volume of water utilized for data center cooling could triple over the next twenty-five years as the demand for raw processing power surges. This disconnect between executive rhetoric and independent forecasting highlights the growing pressure on tech giants to provide transparent metrics regarding their physical resource consumption during the current AI gold rush.

Strategic Rationale and Market Impact

Unlike his total rejection of water-related concerns, Altman conceded that the escalating energy requirements of the industry present a legitimate challenge for global infrastructure. He advocated for a rapid transition toward nuclear, solar, and wind energy to power the next generation of neural networks. According to Altman, the focus should not merely be on the gross consumption of electricity, but rather on the comparative efficiency of AI systems. He suggested that when one considers the twenty years of life and vast amounts of food required to make a human being intelligent, AI models may already be reaching a state of relative energy parity.