Linking the digital world to environmental impacts
Our digital carbon footprint is measured by emissions caused by the production, use and data transfer of digital devices and services. Every online search, e-mail, and streamed video contributes towards the footprint, and the emissions in this context are generally formed in three areas (8):
- storing and processing data in data centers (including cloud services)
- transmitting data through telecommunications networks (including both mobile and fixed-line infrastructure)
- displaying and processing content and data on end-user devices
The environmental impact and the greenhouse gas emissions in these areas are caused by electricity consumption, the manufacturing process of devices and hardware, and water use and consumption for cooling data centers (=servers). Additionally, AI-driven technologies are demanding a massive increase in computing power across the entire digital ecosystem, accelerating the impacts rapidly.
Why data centers?
Data centers are in the center of the impacts. Everything we do in a digital world is mostly taking place within them.
– Alex de Vries, researcher, VU Amsterdam and founder of Digiconomist, on APFI’s webinar series “Backstage of Tech”, 13.3.2025.
To uncover the environmental impact of digital infrastructure, we need to look at data centers. These large facilities house high-performance computing equipment that consume vast amount of electricity and produce substantial waste heat, often requiring water-based cooling methods.
Data centers are now also powering the growing processing and energy demand of AI-driven technologies. In addition to sustaining the operating of AI models, de Vries stresses how the initial (training) stage of AI models is by far more energy and water consuming. What makes it so energy intensive is the large clusters of specialized equipment, and the energy intensive computational power needed to train AI model networks with up to trillions of parameters. The more parameters are used in training, the more robust the training stage becomes, which in turn produces a better AI model, but also increases environmental impact.
In Finland, data centers benefit from access to low-carbon electricity and efficient water-cooling systems, helping reduce their local environmental impact. But the digital ecosystem is global, and the impacts extend far beyond single data center locations. (9)
Many countries are already seeing actual strain on their electricity grids due to data center expansion. Ireland’s data centers, supporting the operations of major companies such as Google, Meta, Microsoft and Amazon, have since 2023 consumed more electricity than the country’s residential homes in total, now standing at 22% of the total energy consumption. By 2028, data centers are projected to consume a third of Ireland’s electricity. (10) In the US, the data center energy consumption will be increasing by 2028 from the current 4.4% of the national electricity consumption to up to 12% (11). De Vries compared, “OpenAI are in 2025 lobbying to open up 5–7 data centers in the USA, which each hyperscale data center consuming up to 40–50 TWh per year, which is roughly half of Finland’s total annual electricity usage.”
The water consumption of data centers is steadily growing as data processing becomes more energy intensive. Shaolei Ren highlights: “Just one tech company consumed 24 billion liters of water in 2023 for its data centers.” This is equivalent to 44% of the total annual residential water consumption in the Helsinki Metropolitan Area, which was 55 billion liters in 2024 (12). According to Ren, even in Finland where the energy grid is relatively low-carbon and efficient cooling systems are used, 24 text-prompt- LLM-responses will consume the equivalent of 0,5L water. And as AI-driven technologies drive up energy use in data centers, there is also a growing public health concern due to emissions and air pollution, both locally and globally, with new scientific research uncovering even more aspects of data center emissions (13).