The surge in artificial intelligence (AI) is dramatically increasing energy consumption, leading to growing concerns about data center emissions. A recent analysis reveals that, between 2020 and 2022, the actual emissions from the in-house data centers of major tech companies—Google, Microsoft, Meta, and Apple—were likely 7.62 times higher than officially reported. Amazon, the largest emitter among the big five tech firms, was excluded from the analysis due to the complexity of isolating its data center-specific emissions, but its total emissions are significantly higher than its peers.
Data centers, which power everything from cloud computing to AI-driven applications, are crucial for the tech industry but consume vast amounts of electricity. According to the International Energy Agency, data centers accounted for 1% to 1.5% of global electricity consumption in 2022, before AI’s rapid growth. AI workloads, such as ChatGPT, are far more energy-intensive than typical cloud-based services. For example, a single ChatGPT query requires about ten times more electricity than a Google search. By 2030, data center energy demand is expected to increase by 160%, with data center emissions potentially reaching 2.5 billion metric tons of CO2 equivalent.
Despite this, major tech companies claim they are achieving carbon neutrality. However, these claims are largely based on “market-based emissions” accounting, which can be misleading. Market-based emissions allow companies to report lower carbon footprints by purchasing Renewable Energy Certificates (RECs). RECs are credits companies buy to claim they are supporting renewable energy production, even if the renewable energy is not consumed directly by their facilities. For instance, a company could be running its data centers on fossil fuels while purchasing RECs from renewable energy sources located far away, making it appear as though their operations are greener than they actually are.
In contrast, “location-based emissions” reflect the actual carbon emissions produced by the energy consumed on-site at data centers. This method accounts for the real environmental impact of the electricity used to power operations, without factoring in RECs or carbon offsets. Many experts argue that location-based accounting provides a more accurate and honest representation of the energy a company consumes and the emissions it generates.
The differences between market-based and location-based data center emissions are striking. Meta, for example, reported its 2022 data center emissions as 273 metric tons of CO2 equivalent using market-based accounting. However, when using the location-based method, that figure jumps to over 3.8 million metric tons—a 19,000-fold increase. Similarly, Microsoft’s market-based data center emissions for 2022 were reported as 280,782 metric tons, but using location-based accounting, the number rises to 6.1 million metric tons—nearly 22 times higher.
These disparities demonstrate how market-based emissions reporting can significantly downplay the environmental impact of data center emissions. Both Meta and Microsoft have substantial gaps between their market-based and location-based emissions because the true emissions associated with their data centers are underreported. The majority of the discrepancy stems from the emissions generated by the energy used to power the data centers themselves, which is mostly ignored in the market-based figures.
In response to these concerns, some tech companies are moving toward more transparent emissions reporting. Google, for example, has adopted a goal to run its operations on renewable energy 24 hours a day, seven days a week, by 2030. This “24/7” model aims to match energy consumption with local renewable energy production, addressing the flaws of the REC system. Similarly, Microsoft’s “100/100/0” initiative seeks to ensure that all of its facilities run on 100% carbon-free energy at all times by 2030.
However, even as these initiatives take shape, the growing energy demands of AI will continue to pose challenges. Experts warn that the data center industry could soon face power shortages, as existing energy infrastructure struggles to keep up with demand. As AI-driven technologies proliferate, tech companies will need to prioritize more accurate reporting of data center emissions and work toward reducing their true carbon footprint, using location-based metrics that reflect the real environmental costs of their operations. Only then can meaningful progress be made in addressing the significant environmental impact of the tech sector’s data centers.
https://www.theguardian.com/technology/2024/sep/15/data-center-gas-emissions-tech