2024-07-17 09:30:44
Google’s greenhouse gas emissions have nearly doubled in 5 years. Microsoft’s have increased by a third in 4 years. Main reasons: the use of artificial intelligence and the increase in the number of data centers. Carbon neutrality targets will be difficult to achieve…
Google is overheating. In its rapport In its annual environmental report, published on July 2, its parent company Alphabet Inc. indicated that its greenhouse gas emissions increased by 48% compared to the 2019 baseline, for a total of 14.3 million tons of carbon dioxide. With a 13% increase in 2023 compared to 2022, its emissions have only increased, while the American giant had set a climate target ” ambitious “ reduction.
Specifically, the report states that by 2023, “Google’s data centers consumed 24 TWh of electricity, or 7-10% of the estimated 240-340 TWh consumed by data centers worldwide, and less than 0.1% of the estimated 25,000 TWh of total global electricity.”.
Google isn’t alone in seeing its GHG emissions soar. In a sustainability report released last May, Microsoft Corp said its emissions were up 29% from a 2020 baseline.
Energy gluttons
The report comes as the International Energy Agency (IEA) has predicted that electricity demand might double by 2026 due to exponential growth in the load of datacenters.
The U.S. Energy Information Administration, which launched a pilot project to track data centers in 2018, found that data centers use 10 to 40 times more energy per square foot than the typical office building and account for nearly 2% of U.S. electricity consumption.
Overall, the IEA estimates that global electricity consumption by data centers is 240-340 TWh, or regarding 1-1.3% of global final electricity demand (regarding 25,000 TWh).
The increase in emissions has highlighted “the challenge of reducing emissions as computing intensity increases and we increase our investments in technical infrastructure to support this AI transition”we can read in the report of Alphabet Inc.
To illustrate how much more demanding AI models are than traditional computing systems, the IEA estimates that a Google search requires an average of 0.3 watt-hours of electricity, while a query on ChatGPT typically consumes around 2.9 watt-hours.
Internet traffic explodes
According to one study published last October by Dutch researcher Alex de Vries (founder of Digiconomist and PhD student at the Vrije Universiteit Amsterdam), Google’s AI systems might, in the worst case, consume as much electricity each year as Ireland, assuming a large-scale adoption of AI in their current hardware and computing systems.
However, the situation does not seem to be improving. According to the latest analyses by international real estate consultancy Savills, the installed capacity of European data centers should reach 13,100 MW in 2027, an increase of 21% compared to current capacity.
Internet traffic in Europe is expected to grow at a compound annual growth rate of 31% by 2030. The European AI market is expected to grow rapidly at 15.9% per year and play a leading role in the increase in demand for data centers.
How can we respond to this explosion of uses (whose effectiveness remains to be proven) and significantly limit GHG emissions? For now, tech giants are focused on responding to this demand.
“For Google, reducing the consumption of its data centers is not its priority. It does not invest in this area, because it has other priorities and because no legislation slows down its expansion by prohibiting the construction of a data center consuming so much or so much energy”explains Ophélie Coelho, independent researcher, specialist in digital geopolitics.
1721266917
#Carbon #neutrality #remains #major #challenge #Google
