New research from the United Nations (UN) has revealed that indirect carbon emissions from Amazon, Microsoft, Alphabet, and Meta have surged by an average of 150% from 2020 to 2023. This increase is attributed to the substantial energy requirements of data centres essential for AI operations, as detailed by the International Telecommunication Union (ITU), the UN agency for digital technologies. Indirect emissions stem from purchased electricity, steam, heating, and cooling used by companies.
Amazon saw the largest increase, with its operational emissions rising 182% in 2023 compared to 2020. Microsoft experienced a 155% increase, while Meta and Alphabet recorded rises of 145% and 138%, respectively. The ITU’s study examined the greenhouse gas emissions of 200 leading digital firms over the same period. Meta, which owns platforms like Facebook and WhatsApp, pointed to its sustainability report, which outlines efforts to cut emissions and resource use at its data centres.
Amazon highlighted its commitment to sustainability through investments in carbon-free energy projects, including nuclear and renewable sources. Microsoft noted improvements in energy efficiency and a shift to chip-level liquid cooling systems to reduce data centre energy consumption. Other companies did not provide immediate comments.
AI investment predicted to escalate global energy consumption
The report forecasts that as investment in AI grows, emissions from the most polluting AI systems could reach 102.6 million tons of carbon dioxide equivalent annually. The energy demands of data centres required for AI development could strain existing infrastructure.
“The rapid growth of AI is driving a sharp rise in global electricity demand, with electricity use by data centres increasing four times faster than the overall rise in electricity consumption, stated the report. “The digital companies assessed in this year’s report consumed an estimated 581 TWh of electricity (2.1% of global demand) with AI being one probable driver of this growth.”
“To deliver on their part, digital companies must adopt absolute, time-bound emission targets across all scopes and fully disclose all relevant Scope 3 emissions, especially in hardware and electronics. Governments and investors should incentivise third-party verification and comprehensive climate reporting in accordance with mandatory corporate GHG reporting schemes.”
Last month, research by Alex de Vries-Gao, founder of Digiconomist, published in the journal Joule, suggests that AI systems could soon account for nearly half of data centre power consumption. This underscores the increasing energy demands of AI technologies, affecting data centre operations globally.
Additionally, an April 2025 report by the International Energy Agency (IEA) predicts that AI will significantly impact the energy sector over the next decade. The report, “Energy and AI,” anticipates that by 2030, data centres will consume over 945 terawatt-hours (TWh) of electricity, exceeding Japan’s current total electricity consumption, with AI-enhanced data centres expected to quadruple their electricity demand.
Read more: AI to propel global data centre electricity demand by 2030
