Sam Altman: Public Concerns Over AI Data Center Water Usage Are Exaggerated
OpenAI CEO Sam Altman stated that public anxiety regarding AI data center water consumption is an overblown myth unsupported by real data. He highlighted metrics showing an average ChatGPT query consumes only about 0.32 mL of water, placing data centers' resource footprint on par with standard office buildings. As public opposition to data center construction rises due to environmental concerns, tech leaders are actively countering with resource-efficiency metrics. Clarifying the actual water impact helps separate public perception from measurable environmental resource consumption as AI infrastructure continues to scale. Altman contrasted ChatGPT's 0.32 mL per query usage (and Google Gemini's 0.26 mL) with agriculture, noting that growing a single California almond requires 3.56 liters of water—more than thousands of AI queries. A 2024 Virginia state commission report similarly found that most data centers consume water at rates comparable to or lower than standard large office buildings.
## BACKGROUND
Data centers generate substantial heat from high-density computing servers and traditionally rely on evaporative cooling systems alongside air cooling to maintain optimal operating temperatures. Evaporative cooling can lose up to 85% of consumed water into the atmosphere, making localized resource draw a sensitive issue for surrounding communities. Industry metrics like Water Usage Effectiveness (WUE) measure how efficiently data centers utilize water relative to their energy consumption.