UN General Assembly highlights AI and climate nexus amid global challenges
At last week's United Nations General Assembly, Secretary-General António Guterres identified three existential threats facing humanity: climate change, unchecked artificial intelligence and widening income inequality. While these challenges are complex and interconnected, with varying levels of concern among global powers, they are not insurmountable.
In an era of rapid narrative shifts driven by social media, public discourse has moved swiftly from catastrophizing about climate apocalypse to forecasting the myriad ways AI will damage society and the environment. Yet widespread despair has never produced constructive outcomes.
Climate Week NYC, which ran concurrently with the General Assembly, offered more nuanced perspectives. Climate and AI topped the agenda there as well, but with less zero-sum messaging and more focus on actionable solutions.
The environmental cost of AI infrastructure
AI's environmental impact is substantial and increasingly difficult to justify given the scale of the climate crisis and persistent failure to reduce greenhouse gas emissions. Data centers powering AI consume scarce land, vast water supplies and enormous amounts of energy, still predominantly derived from fossil fuels.
The International Energy Agency projects global data center electricity consumption will reach 1,000 TWh by 2026, more than double the 2022 level, driven largely by AI workloads. By 2030, 40 percent of electricity used by data centers will support AI operations, equivalent to the residential electricity needs of all 1.3 billion people in sub-Saharan Africa, according to technology insight company Gartner. This comparison carries particular weight given that over 600 million people in the region still lack electricity access entirely.
Bloomberg New Energy Finance estimates AI's load on data centers will increase global power sector emissions by 10 percent over the coming decade. This comes on top of emissions from other computing activities linked to cloud storage, video streaming and cryptocurrency mining.
The scale of resource consumption extends beyond electricity. Research published in Nature indicates that water consumption by United States data centers is projected to reach 1.7 billion gallons per day by 2027, equivalent to the daily water needs of approximately 6.4 million people. Training a single large language model can emit roughly 300 tons of carbon dioxide, as much as five cars over their entire lifetimes, according to University of Massachusetts Amherst research.
A critical juncture for climate action
The AI debate intensifies as scientists confirm that 2024 marked the first year global average temperatures exceeded the 1.5 degrees Celsius threshold above pre-industrial levels, reaching 1.54°C according to the World Meteorological Organization. The more evident the climate crisis becomes, the more questionable it appears to embrace a technology with an insatiable appetite for electricity and water.
Yet AI is here to stay, and plans to colonize Mars remain distant. At Climate Week NYC, greater attention was paid to mounting evidence that AI can help tackle the climate crisis. Applications include intelligent load management for renewable energy, more effective pollution tracking by integrating disparate data from weather stations, satellites and private monitors, and improved climate change modeling for adaptation and mitigation. Research from Google DeepMind and the European Centre for Medium-Range Weather Forecasts shows AI-powered climate modeling has improved extreme weather prediction accuracy by 20-30 percent compared to traditional methods.
There was also recognition that as the United States rolled back climate commitments at the federal level, Big Tech companies were playing an important role in funding renewable energy projects through multimillion-dollar contracts. Major technology companies including Microsoft, Google and Amazon have signed power purchase agreements totaling over 50 GW of renewable energy capacity as of 2025, representing approximately 60 percent of all corporate renewable energy procurement globally.
Beyond greenwashing narratives
This more balanced perspective on AI's climate potential should not enable corporate greenwashing. While hopes exist that data centers will increasingly run on renewable sources, the immediate effect of bringing coal or gas-fueled power plants online to meet current energy demands is severely damaging, undermining years of work to retire fossil fuel generation.
Ketan Joshi, a clean energy and corporate accountability researcher, highlighted in a recent paper that many claims about AI's climate benefits link back to case studies published by conflicted corporations rather than independent academic research. Joshi also notes that AI benefiting the climate typically involves traditional machine learning, which has far lower environmental impact than generative AI tools like Copilot, Gemini or ChatGPT. A single ChatGPT query can consume up to 10 times more electricity than a traditional Google search, according to the Electric Power Research Institute. With AI-related queries now accounting for approximately 15 percent of total searches on Google's platform as of early 2026, the cumulative impact is significant.
Responsibility and regulation
This awareness creates pressure on consumers to make ethical choices about when and how they use AI. Since generative AI carries greater environmental cost, users should be better informed about when they genuinely need extra computational power. Guidelines suggest using traditional search engines for simple queries like recipes, avoiding unnecessary image and video generation, and declining automatic AI search options.
However, placing the burden solely on end users is inadequate. The climate dimension must be central to ongoing AI regulation and sovereignty debates. The EU's AI Act, which entered into force in August 2024, includes provisions requiring high-risk AI systems to meet energy efficiency and environmental sustainability standards, offering one regulatory model.
As with recent political discourse, these issues must not be hijacked by zero-sum thinking, knee-jerk polarization or absolutism. The dire warnings punctuated by glimpses of hope emerging from New York last week remind us that progress on climate, AI and related challenges depends on empirical evidence, nuance, transparency and accountability.
