COP30 is underway in Belém, Brazil, and despite the United States’ absence and concerns about a U.S. retreat from global climate engagement, the summit has already been eventful. Demonstrators and Indigenous leaders interrupted the proceedings multiple times to protest environmental damage in their communities. Twelve countries, including Canada, signed on to a new declaration to address climate-related disinformation. And several countries are forming a coalition to outline a roadmap for moving away from fossil fuels.

Amid these headline developments, artificial intelligence was everywhere at COP30, though not always in expected ways. Attendees were invited to use “What’s UP, Belém?”—a multilingual AI agent designed to support “responsible, authentic tourism.” The summit also features an AI chatbot named Macaozinho, represented by a cartoon Brazilian macaw and trained on UN documents to act as an “ally in the midst of an environment of fake climate news.” The estimated water and energy consumption behind these tools was not disclosed.

Related Story: Artificial Intelligence Sparks Debate at COP30 Climate Talks

More substantially, the summit marked the launch of a new AI Climate Institute, designed to “equip people and institutions in developing countries with crucial skills to harness AI for climate action.” Official communications framed it as a “tool of empowerment.” Earlier sessions also showcased a number of technological success stories, including Brazilian advances in forecasting and responding to extreme weather. 

Summit documentation consistently refers to technology as a climate “ally.” A new Green Digital Action Hub articulates a vision to support climate action through digital technologies, with a strong focus on the Global South, but also to “reduce the environmental footprint of technology.” Winners of the AI Environmental Footprint Measurement Hackathon were also announced at COP30, though their findings have not yet been released.

The sense of optimism around using digital tools to spur climate and resilience efforts is not misplaced. Around the world, digital tools and AI are helping predict extreme weather events, improve building operations, and monitor biodiversity and conservation efforts. But COP30 discussions have paid relatively little attention to the environmental costs of adopting these systems: even as AI supports climate action, it drives higher demand for water, energy, and the natural resources required to build and operate its hardware.     

Brazil, the COP30 host, has been competing to attract data centres while managing concerns about their impacts. One new data centre planned by the parent company of TikTok is targeting a location near Indigenous territory, while another is planned for a region devastated by extreme flooding last year.

AI investments continue to grow at an astounding rate, with a recent report estimating that over US$5 trillion will be spent on data centres and related power supply over the next 5 years. Bubble concerns aside, it’s clear that this industry will be a powerful global influence in the years ahead. 

Understanding the environmental implications of AI is not straightforward, with new and often conflicting reports on its energy use emerging each week. Although we may lack the desired clarity and precision about AI’s climate risks versus its benefits, we know enough to recognize that global leadership must turn its focus to this question. Given the disproportionate costs borne by residents of marginalized communities and the Global South, ensuring safeguards and energy measurement in the AI industry must be a central topic in global climate discussions.