Pope Leo XIV: Transitioning from Climate Crisis Response to Prevention in the Age of AI

Data centers hosting climate prediction models consume enough energy daily to power tens of thousands of households and use significant amounts of water for cooling. As neural networks like GraphCast or Pangu-Weather become more complex and machine learning relies on larger volumes of satellite data, this pressure increases. In response, Pope Leo XIV has issued a unprecedented call: instead of focusing solely on emergency responses to climate disasters, he urges governments and scientists to prioritize prevention in public policies. His speech, delivered at a Vatican assembly on climate, highlights that the excessive energy consumption of AI threatens essential natural resources crucial for societal resilience.

Announcement Details: Who, What, Precise Figures

Pope Leo XIV, during a public audience, stated that artificial intelligence systems, particularly large-scale predictive models, "exert unprecedented pressure on natural resources" due to their high energy and water consumption. He noted that global data centers powering these models are increasingly contributing to global energy demand, although no exact figures were provided in the official statement. This warning is part of a letter addressed to heads of state, climate researchers, and tech industry leaders.

In the same text, Pope Leo urged climate actors to shift from an "emergency response" approach, often limited to managing immediate disasters, to a proactive prevention strategy. He emphasized the importance of investing in green infrastructure, reducing forecast uncertainty with more precise atmospheric data, and promoting low-carbon footprint AI models to mitigate future warming. No specific figures were given, but the message emphasizes the need for rapid and coordinated action.

Context: Climate Objectives, Current Emission Situation, and AI

The Paris Agreement aims to keep global warming well below 2°C, with efforts to limit it to 1.5°C. To achieve this, institutions like ECMWF and the Copernicus program continually improve their weather models by integrating high-resolution satellite data to reduce forecast uncertainty. However, the proliferation of machine learning models using deep neural networks increases energy demand in the sector. Latest estimates suggest that training large climate prediction models can consume as much energy as several hundred households for a year, although the exact figure was not specified in the Vatican report.