The Environmental Impact of Advances in Artificial Intelligence

Artificial intelligence (AI) is transforming industries globally. At the same time, it poses significant environmental challenges. Technology leaders like Google, Microsoft, and Amazon must navigate between innovation and environmental responsibility.

The environmental challenges of data centers

Data centers are essential for AI, but operating them consumes significant amounts of energy and water, increasing carbon emissions. In 2022 and 2023, these centers, with cryptocurrencies and theAI, used 460 TWh of electricity, or 2% of global demand, and up to 4% in regions densely populated by data centers such as the US, China and the EU. With around 10,000 data centers around the world, including 33% in the United States, 16% in Europe and 10% in China, their impact on energy consumption is considerable, representing 1.5% to 2% of global greenhouse gas emissions.

Focus on the major players in AI

Google emissions have thus increased by 48% since 2019, with a similar increase being observed at Microsoft and Amazon. According to Microsoft's 2023 Sustainability Report, its carbon footprint grew by 29% in one year, mainly due to new data centers. The construction of these facilities has become a main cause of the increase in CO2 emissions associated with generative AI, underlining the urgency of developing more sustainable practices.

Expansion and impact of new data centers

The forecasts of the International Energy Agency (IEA) estimate that the power consumption of data centers could be between 620 and 1,050 TWh in 2026, with a base estimate of more than 800 TWh, marking a significant increase from 160 to 590 TWh since 2022. This rapid growth is accompanied by a continuous expansion of infrastructure. In Europe, the number of data centers is expected to more than double by 2025 to meet the growing demand for storage. In the United States, construction projects planned until the second half of 2023 will already exceed existing capacity. This expansion highlights the need to develop infrastructures that have a lower environmental impact.

In search of sustainable strategies

To meet environmental challenges, companies like Microsoft and Amazon are investing in regions like Aragon in Spain, benefiting from renewable energy resources. These data centers, designed to be powered by solar and wind energy, represent an initiative to reduce ecological impact while supporting growing digital demand. Nevertheless, although the increase in electricity demand is anticipated, greenhouse gas emissions are expected to remain stable at around 2% of global emissions, thanks to the acquisition of Green PPA by the main data center operators.

Regulation and future perspectives

Although data centers have improved their energy efficiency, this is still insufficient in the face of the rapid growth in demand for computing power. The regulation of generative AI is progressing slowly: the European AI Act voted in March 2024 will not be effective until 2026. Businesses must therefore self-regulate without waiting for these regulations, setting their own limits to minimize reputational and financial risks. Using existing buildings for data centers rather than building new ones could also reduce environmental impact.

Conclusion

Advances in AI offer numerous benefits but also pose environmental risks. For innovation to be truly beneficial, technology companies must adopt an environmentally conscious approach, integrating sustainability at the heart of their innovations. Each development must be evaluated both for its technological potential and for its environmental impact, in order to promote a more sustainable future.

‍