The report stresses that governments and industry should invest in energy-efficient AI research, development, and literacy to mitigate environmental impacts and promote sustainable growth.

Most AI infrastructure is concentrated in high-income countries, with only about 5% of Africa’s AI talent having access to the necessary computing power, which underscores the need for more accessible, resource-efficient AI solutions globally.

Implementing smaller and more efficient AI models can help bridge global inequalities by making AI technology more accessible in low-resource settings, especially in regions like Africa where infrastructure is limited.

The report advocates shifting away from resource-heavy models toward efficient, specialized, and compressed AI models to support sustainable development worldwide.

Using smaller, specialized models and optimized prompts not only conserves energy but also improves accessibility in low-resource areas with limited connectivity and energy supplies.

These energy-efficient models require less computing power, making AI faster and more sustainable, which helps reduce environmental impacts.

Currently, each AI request, such as those made to ChatGPT, consumes about 0.34 Wh, with daily requests totaling roughly 310 GWh annually—comparable to the electricity consumption of three million people in Ethiopia.

Generative AI tools like ChatGPT, Gemini, and Copilot are used by over 1 billion people daily, and their substantial computational demands highlight the urgent need for more energy-efficient AI solutions.

A recent UNESCO report highlights that reducing user query lengths and employing smaller, specialized AI models can cut AI energy consumption by up to 90% without sacrificing performance, addressing the growing environmental concerns associated with AI.

AI energy demand is doubling approximately every 100 days as generative AI becomes more embedded in daily life, putting immense strain on global resources like energy, water, and minerals.