Are New Technologies (AI, Data Centers) Compatible with Environmental Sustainability Goals?

The growing importance of cryptocurrencies, artificial intelligence (AI), and data centers in the global economy, as well as their impact on daily life, raises crucial questions about their ecological footprint and sustainability. The International Energy Agency (IEA) estimates that these technologies consumed around 460 TWh of electricity worldwide in 2022, accounting for nearly 2% of total global electricity demand. While this consumption may seem negligible compared to other sectors such as transportation, the IEA warns in its report Electricity 2024. Analysis and forecast to 2026 on the risk of doubling electricity demand in the aforementioned sectors by 2026. Although forecasting in the sector of new technologies (data centers, cryptocurrencies, and AI) is difficult to grasp due to the pace of their deployment and rapid evolution, especially in terms of energy efficiency, the risk of increased electricity demand raises several questions regarding the ability of energy systems to maintain a balance between supply and demand. It also raises concerns about how this energy is produced and distributed across the territory. In other words, are these new technologies compatible with the environmental sustainability goals set by the Paris Agreement?

An increasing electricity demand for new technologies by 2026

The IEA presents three scenarios projecting electricity demand for the data center and artificial intelligence (AI) sectors over the next five years. The global electricity consumption for these sectors in 2026 is expected to range between 620 TWh for the most efficient scenario and 1,050 TWh for the most energy-intensive scenario, equivalent to the current electricity consumption of Japan. The intermediate scenario estimates a demand around 800 TWh in 2026, compared to 460 TWh in 2019. According to the projections, the demand could increase by between 160 TWh and 590 TWh in 2026 compared to its 2022 level, which would be equivalent to adding, at a minimum, Sweden’s electricity consumption, or even that of Germany for the high-end scenario.

Data centers are a central component of digital infrastructure and are significant contributors to the growing electricity demand. They alone account for between 1% and 1.3% of global electricity consumption, or between 260 and 360 TWh. This increase is driven by the growing production of digital data, which necessitates larger data centers to store and process this data in the cloud. The rapid integration of AI into software programming across various sectors (deep learning, cloud computing) is further driving the global electricity demand for servers. Training AI models, particularly in deep learning, requires large amounts of data and, consequently, substantial computing power. According to the IEA report, search engines like Google could see their electricity demand surge by 2026 with the full implementation of AI. By comparing the average electricity demand of a typical Google search (0.3 Wh) to the 9 billion daily queries of ChatGPT, the AI developed by OpenAI (2.9 Wh per query), the IEA estimates the additional annual electricity consumption at 10 TWh.

The strategy of territorializing energy infrastructures

The expansion strategies of tech multinationals involve the territorialization of their infrastructures. While local communities view the arrival of these giants as a significant financial resource with many jobs at stake, these firms engage regional resources such as water, electricity, and components over the long term. In this context, some countries are questioning the robustness of their electrical systems and are imposing restrictions on the establishment of additional data centers in their territories: this is the case in Ireland, Germany, Singapore, China, and some states in the United States. For example, the electricity network operator for the city of London, National Grid Electricity Transmission, announced in 2022 that it would no longer allow new data centers to connect to the grid, making the construction of new housing difficult, if not impossible, without drastic investment in the electrical network. The same situation applies to the Irish capital, Dublin, where the public electricity transmission operator, EirGrid, halted several data center construction projects for supply security reasons, including two led by Amazon Web Services and Microsoft. Attracted by significant tax breaks, data server construction projects are numerous, and their electricity consumption is expected to represent between 23 and 30% of national consumption by 2030, according to EirGrid’s estimates.

Although it is difficult to predict the precise evolution of data storage demand, it is now certain that new technologies will play an increasingly important role in our daily lives. While there are solutions to limit the electricity consumption of these technologies, it is essential to first question the carbon footprint of our virtual lives and how our societies manage and share regional renewable resources. Solutions based on improving energy efficiency or decarbonizing the energy mix exist, but they overlook the issue of fair distribution of natural resources between territories, on the one hand, and between populations and industries, on the other.

---

See Ophélie Coelho, Géopolitique du numérique. L’impérialisme à pas de géants. (Paris : Les Éditions ouvrières, 2023)