Overview
- An analysis of 154 corporate statements found that only 26% cited published academic research and 36% offered no evidence for claimed climate benefits.
- The report says companies conflate lower-impact traditional AI with energy-intensive generative AI, creating a bait‑and‑switch that overstates climate gains.
- A widely repeated claim that AI could mitigate 5–10% of global emissions by 2030 traces to a 2021 BCG blog post and a Google‑commissioned report based on consultant experience.
- Independent forecasts point to rising energy demand: the IEA projects data center consumption will double by 2030, and BloombergNEF estimates the U.S. share could reach 8.6% of electricity use by 2035.
- Google defended its methodology as grounded in the best available science; Microsoft declined to comment, and the findings were released at the AI Impact Summit in Delhi.