Sam Altman: AI’s energy appetite, not water use, is the real concern as data centres expand

OpenAI CEO Sam Altman defended modern AI data centres, calling online claims about large water use "completely untrue" and saying newer facilities avoid water cooling. He acknowledged energy consumption as the pressing issue and urged a rapid shift to nuclear, wind and solar as AI demand grows. Altman compared the costly training phase of models to raising humans for 20 years while arguing energy efficiency should be measured per query after training. Indian tech billionaire Sridhar Vembu criticized equating technology with humans, reflecting broader unease about AI replacing jobs. The article highlights the US trend toward massive off-grid data centre projects — some planning to run primarily on natural gas plus solar — backed by Meta, OpenAI, Oracle and others. Examples include an 8,000-acre West Texas project and multiple sites in Wyoming, New Mexico, Pennsylvania, Utah, Ohio and Tennessee. Analysts warn off-grid gas-powered centres threaten climate goals; the EPA has enforced emissions rules against at least one setup (Elon Musk’s Memphis facility). Data centres consumed electricity comparable to entire countries (e.g., Germany or France) after the 2022 AI boom. The US runs 5,246 data centres using at least 17 GW; many off-grid projects are planned, and some states have eased approvals. Local opposition is rising over water use, emissions and higher local electricity costs; PJM grid prices have jumped and tech firms pledged $15 billion toward new generation capacity. Key points for traders: accelerating AI-driven power demand may shift capital toward energy and infrastructure equities, increase regulatory scrutiny, and create localized grid stress that can affect data-related crypto services and hosting costs.