The expansion of artificial intelligence (AI) is accelerating investment in data centers worldwide, but it brings with it a set of environmental costs that remain largely outside the public debate. The conclusion comes from a new Allianz Trade report, which warns of the sector’s growing impact on electricity consumption, carbon emissions, and the use of water resources.
According to the analysis, global investment in data centers is expected to reach $580 billion in 2025, in a context marked by the rapid spread of AI applications. The installed capacity of these centers is expected to double by the end of the decade, accompanying an unprecedented energy demand.
Allianz Trade estimates that the global electricity consumption of data centers will reach about 515 TWh in 2025, a figure that could more than double to 1,111 TWh by 2030. A growing portion of this demand is linked to artificial intelligence: currently, AI-related workloads account for between 15% and 20% of the electricity consumed by the sector, a share that could approach 40% within five years.
The report also highlights that the carbon footprint of data centers is often underestimated. By including not only the electricity used in operations but also the emissions associated with the construction of infrastructure, hardware manufacturing and grid losses, Allianz Trade calculates global emissions of 286 million tonnes of carbon dioxide (MtCOâ‚‚) in 2025. This figure is about 57% higher than some international estimates focused solely on operational consumption.
The United States and China account for approximately 70% of global data center emissions, reflecting the strong concentration of digital infrastructure in these two markets. However, environmental impact varies significantly depending on geographic location. The same AI task can generate up to 24 times more emissions depending on the country where it is processed, due to differences in the carbon intensity of electricity grids.
Countries such as Indonesia, India and Malaysia exhibit emission factors above 600 grams of COâ‚‚ per kilowatt-hour, while Norway and Sweden record values below 30 grams, thanks to the high integration of renewable energy.
Europe appears relatively well positioned on this indicator. Although the European countries analyzed represent around 15% of the global data center capacity, they account for only 8% of the indirect emissions associated with electricity consumption, benefiting from an energy mix less dependent on fossil fuels.
The sector’s future will depend largely on the speed of decarbonization of electric systems. In a scenario where the carbon intensity of grids does not improve, global data center emissions could reach 643 MtCOâ‚‚ in 2030. In a path aligned with more ambitious climate goals, that figure could sit around 329 MtCOâ‚‚.
The economic dimension of these impacts is also significant. The study estimates that the climate costs associated with the sector will amount to about $68 billion per year in 2025. Without progress in decarbonizing electricity, this amount could rise to $154 billion by the end of the decade.
Artificial intelligence represents an increasing slice of this environmental bill. Currently, AI applications are responsible for about $13 billion in annual climate costs, a figure that could exceed $50 billion by 2030, depending on technological and energy developments.
Beyond energy and emissions, water consumption emerges as another concern. Data centers will have consumed about 814 billion liters of water in 2025, mainly for cooling systems. By 2030, consumption could range between 1.3 and 1.8 trillion liters, a volume comparable to Switzerland’s annual water usage.
The impact will not be uniform. Allianz Trade identifies markets such as South Korea, India, Mexico, Spain, and China as regions where accelerated data center expansion could exacerbate pressure on water resources already subject to high stress.
Despite the challenges, the report emphasizes that AI can also play a meaningful role in the energy transition. Citing estimates from the International Energy Agency (IEA), Allianz Trade notes that applying artificial intelligence solutions in the energy, industry, buildings and transport sectors could enable global emissions reductions of about 1.4 gigatonnes of COâ‚‚ per year by 2035, provided it is implemented at scale.
To minimize the environmental impacts of the digital revolution, the insurer advocates a strategy based on expanding renewable energies, strengthening electrical grids, investing in storage and energy flexibility, greater transparency about energy and water consumption, and reducing embodied emissions in equipment and technological infrastructures.
As artificial intelligence becomes a central component of the global economy, the report concludes that the sustainability of digital infrastructure will be one of the decisive factors in determining whether the benefits of the technology outweigh its environmental costs.
