The rapid rise of artificial intelligence (AI) has drastically increased energy demands for big tech companies, especially in their data centers. While companies like Google, Microsoft, Meta, and Apple claim carbon neutrality, a Guardian analysis suggests that from 2020 to 2022, their real emissions from in-house data centers were likely 662% higher than officially reported. Amazon, the largest emitter among the tech giants, was excluded from the analysis due to its different business model, but its carbon footprint remains significant.
Data centers already account for 1% to 1.5% of global electricity consumption, a figure expected to grow as AI systems like ChatGPT become more prevalent. AI is highly energy-intensive, with a single ChatGPT query requiring nearly 10 times more electricity than a typical Google search.
Much of the reported emissions reductions are attributed to the use of renewable energy certificates (Recs), which allow companies to purchase certificates representing renewable energy consumption, even if the energy is generated far from the company’s operations. This leads to discrepancies between official market-based emissions and more accurate location-based emissions, which reflect the actual carbon emissions from the area where the data centers operate.
If the five big tech companies were a country, their 2022 location-based emissions would rank 33rd globally. Meta’s real emissions, for example, jumped from 273 metric tons of CO2 under market-based accounting to more than 3.8 million metric tons when using location-based metrics.
A debate over emissions accounting methods is ongoing between two tech factions. Amazon and Meta support retaining Recs, while Google and Microsoft advocate for a more stringent, location-based approach to match renewable energy production with data center consumption. As tech companies’ energy use continues to rise, these discrepancies highlight the need for more accurate carbon accounting to reflect the environmental impact of data centers.