A new peer-reviewed study finds that artificial intelligence adoption increases net global carbon dioxide emissions by 0.5 to 1.8 gigatonnes annually — equivalent to 1.2% to 4.8% of 2024’s energy-related CO₂ — because the technology accelerates oil and gas production as effectively as it optimizes solar, wind, and grid operations. The research strips away the prevailing narrative that AI is a net climate positive by quantifying how efficiency gains are split roughly evenly between clean and fossil energy systems.
The study’s methodology matters because it captures rebound and enabling effects that narrower analyses miss. When AI reduces the cost of seismic imaging or reservoir modeling, it extends the economic life of marginal oil fields. When it sharpens wind-farm layout or inverter control, it adds renewable megawatts. Both outcomes are real, measurable, and of comparable magnitude. The net result is not decarbonization but a more energy-intensive economy overall.
For policymakers and investors, this reframes the AI-and-climate conversation. Subsidizing generic AI compute or data-center build-out without conditionalities — such as binding clean-power procurement or emissions-intensity thresholds — risks locking in the very emissions the technology is supposed to help abate. The International Energy Agency has already flagged data-center electricity demand as a growing share of global load growth; this study adds a supply-side dimension that is harder to mitigate with procurement alone.
Oil majors have been among the earliest large-scale adopters of industrial AI, using it to cut drilling days, optimize well spacing, and reduce methane slip. Those same toolchains are now being marketed to renewable developers. The technology itself is agnostic; the carbon outcome depends entirely on where capital and policy direct it. Without explicit guardrails, the default is acceleration of the status quo.
Read the full report at Energy Central.