Artificial intelligence and the climate dilemma
https://arab.news/rrsfx
Artificial intelligence is rapidly transforming economics, science, and governance. Yet behind these benefits lies a steadily growing cost: environmental degradation.
AI systems demand enormous amounts of energy and water to operate, including the giant computing clusters, data centers, and cooling systems they rely on. This tension is at the heart of the AI paradox: the very technology that can help mitigate climate change may also worsen it. Depending on its design, power source, and governance, AI can either be a solution or a liability for the planet.
Current AI models are inherently energy-intensive. Training large language models can consume hundreds to thousands of megawatt-hours of electricity. For example, training GPT-3 reportedly required approximately 1,287 MWh of electricity, resulting in over 550 tonnes of carbon emissions when powered by a conventional electricity grid.
Energy demand does not stop at training. Inference — the energy needed to respond to user queries — occurs continuously and grows with global usage. Even minor energy costs per query add up significantly across millions of daily queries. Moreover, the carbon footprint of AI depends heavily on the electricity grid: fossil-fuel-heavy grids produce far more emissions than renewable-based ones.
Data centers, the physical backbone of AI, are another environmental concern. Current research estimates that data centers consume 2-3 percent of the world’s electricity — a figure that could rise to 8-10 percent by 2030 under rapid AI growth.
AI-optimized facilities operate 24/7 and produce excessive heat, requiring intensive cooling. Most rely on evaporative cooling, which consumes millions of liters of freshwater daily. In water-stressed regions like Arizona, Ireland, and parts of the Gulf, this has sparked social and political tensions over whether data infrastructure should take precedence over residential and agricultural needs. In many ways, data centers are the battleground where the AI sustainability debate plays out.
Despite these challenges, AI also holds great potential to advance climate action. AI-based climate models can improve predictions of extreme weather, enabling preventative measures that save lives. Machine learning can optimize renewable energy grids by forecasting demand and balancing intermittent supply. Satellite-based AI systems, such as those developed by Carbon Mapper, detect methane leaks in near real-time, addressing high-impact emissions.
Predictive maintenance and supply chain optimization further reduce waste and support the circular economy. These applications demonstrate that AI’s environmental value can extend far beyond computational efficiency.
The issue is not AI itself but the systems it relies on. Fossil-fuel-powered grids convert computational expansion into emissions, while water-intensive cooling strains local ecosystems. This has given rise to the terms Green AI and Red AI.
Green AI emphasizes efficiency, renewable energy, and resource-friendly design. Red AI prioritizes scale and performance regardless of environmental impact. The climate legacy of AI will depend on which path we choose.
Encouragingly, solutions are emerging. Corporate giants like Google, Microsoft, and Amazon are leading the way with long-term renewable energy contracts. Advanced cooling techniques, including liquid immersion and seawater cooling, can drastically reduce freshwater use. Carbon-aware computing — redistributing workloads to locations and times with lower carbon intensity — is another promising strategy. Governments in the EU, Singapore, and the UAE are developing AI policies to improve energy efficiency, cut emissions, and promote sustainable practices.
AI will not automatically become a climate-positive technology. It must be developed with governance, innovation, and accountability at the forefront. Within planetary limits, AI could become one of humanity’s most powerful tools against climate change. Without care, it risks becoming a major source of environmental stress. The future of AI is not a choice between progress or the planet — it is about advancing in ways that safeguard both.
- Majed Al-Qatari is a sustainability leader, ecological engineer and UN youth ambassador.
