Google's greenhouse gas emissions soar by almost 50% due to AI energy demand

Google's greenhouse gas emissions have increased by 48% or more since 2019, and electricity consumption has increased dramatically as the degree of dependence on AI in data centers of major search companies increases.

Google currently admits in the company's 2024 Environmental Report that it is “not easy” to achieve the “extremely ambitious” goal of achieving net zero emissions by 2030, and that achieving the goal involves “great uncertainty.”

In 2020, Alphabet Ink (NASDAQ: GOOG), the parent company of Google, announced a 24/7 CFE target of operating all businesses with carbon-free energy by 2030 and matching electricity demand with CFE supply every hour every day.

In May, Google revealed plans to invest 1.1 billion dollars to expand its main data center in Finland. However, thanks to the progress of the AI boom, the company will have to contend with the rapid increase in electricity demand seen in various industries in the Western economy.

Last year, the power consulting company Grid Strategies published a report titled “The Era of Flat Electricity Demand Is Over,” and it was revealed that US transmission grid planners, including regional transmission operators (RTOs) and electric power companies, have almost doubled their five-year power demand growth forecasts. AI in particular is driving much of the growth in electricity demand.

According to the Electric Power Research Institute (EPRI), data centers will consume up to 9% of the total power generation in the United States by 2030 from the current ~ 1.5% due to the rapid introduction of technologies that consume large amounts of electricity, such as generative AI. For reference, the US industrial sector's energy consumption last year was 1.02 millionGWh, accounting for 26% of US electricity consumption.

The survey results are hardly surprising given that AI servers are true consumers of electricity. Digiconomist estimates that one NVIDIA DGX A100 server consumes enough electricity for several American homes, while hyperscale data centers consume 1 GW of electricity equivalent to the output of a nuclear power plant.