Resource consumption and carbon footprint of AI technologies
Baranchuk, Aleksandra (2026)
Kandidaatintyö
2026
School of Energy Systems, Ympäristötekniikka
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe20260619100527
https://urn.fi/URN:NBN:fi-fe20260619100527
Tiivistelmä
This thesis examines the resource consumption and carbon footprint of artificial intelligence (AI) infrastructure, tracing the chain from financial investment to physical infrastructure and environmental impact. The study addresses the lack of mandatory reporting for hyperscale data centres in the United States, where AI development is expanding rapidly without regulatory accountability. Through a literature review and comparative study of Europe and the United States, the research investigates how regional and regulatory differences shape energy consumption, water use, and carbon emissions.
The analysis reveals that the same technology produces different environmental outcomes depending on location. In Texas, data centres mostly operate on a fossil fuel-dependant grid, where they benefit from tax incentives that divert revenue from local services. In the European Nordic region, data centres source the majority of their electricity from renewables, as well as utilise colder climates and abundant water for cooling. Data centres in Europe operate under mandatory energy and water reporting, mandated by the EU Energy Efficiency Directive, and recover waste heat for district heating, providing direct community benefit.
The key finding indicates that the difference in carbon footprint depends on regulatory environments rather than technological capability. US deregulation allows large corporations to dominate the infrastructure, while European mandates require sustainability checks, resulting in lower emissions. Europe demonstrates that sustainable AI infrastructure is achievable. Nevertheless, AI remains an industrial-scale driver of climate change, with its carbon footprint among the fastest-growing sectors globally.
The analysis reveals that the same technology produces different environmental outcomes depending on location. In Texas, data centres mostly operate on a fossil fuel-dependant grid, where they benefit from tax incentives that divert revenue from local services. In the European Nordic region, data centres source the majority of their electricity from renewables, as well as utilise colder climates and abundant water for cooling. Data centres in Europe operate under mandatory energy and water reporting, mandated by the EU Energy Efficiency Directive, and recover waste heat for district heating, providing direct community benefit.
The key finding indicates that the difference in carbon footprint depends on regulatory environments rather than technological capability. US deregulation allows large corporations to dominate the infrastructure, while European mandates require sustainability checks, resulting in lower emissions. Europe demonstrates that sustainable AI infrastructure is achievable. Nevertheless, AI remains an industrial-scale driver of climate change, with its carbon footprint among the fastest-growing sectors globally.