AI data centers are expanding fast, but their carbon footprint is difficult to pin down. CDP, the carbon disclosure nonprofit, says the largest US hyperscalers have not published detailed energy or emissions data, so investors and customers cannot compare the climate impact of individual sites. BloombergNEF projects data center electricity demand will quadruple by 2035 and reach 7% of global power consumption. Without site level reporting, decisions about where to buy AI computing power are being made in the dark. CDP is pushing for more granular disclosures, including water use for cooling, and for companies to publish credible transition plans with year on year data instead of distant targets. The stakes are high: one proposed campus in West Virginia would need more than 8 GW, roughly the electricity demand of New York City.
