The rapid development of artificial intelligence could lead to a measurable increase in carbon emissions in the United States, according to a recent study published in the journal Environmental Research Letters. Researchers assessed the impact of widespread AI adoption on electricity consumption and carbon dioxide emissions on a national scale.
According to their estimates, a massive rollout of these technologies could generate about 900,000 additional tonnes of CO₂ per year. While this volume is modest in comparison to the country’s total emissions, it represents a significant and quantifiable contribution to the American carbon footprint.
In terms of energy, the expansion of AI would require up to 12 petajoules of additional electricity each year, equivalent to the annual consumption of about 300,000 American households. These additional needs are primarily driven by the increased computing power required to train and operate artificial intelligence models.
One of the study’s co-authors emphasizes that, although these emissions remain lower than those from many other economic sectors, they need to be anticipated. He advocates for incorporating energy efficiency measures into the design of AI systems to mitigate their long-term environmental impact. Suggested approaches include optimizing algorithms and increasing reliance on low-carbon energy sources.
Data centers, which are central to AI operations, account for most of this energy consumption. These infrastructures, which host servers dedicated to intensive calculations, require significant amounts of electricity to power processors and ensure their cooling. The growth of AI is thus driving the construction of new facilities, which intensifies energy demand, often still largely dependent on fossil fuels.
To address these challenges, engineers are developing more efficient solutions, such as compressing models to reduce computing needs without significantly compromising performance, or utilizing specialized processors like tensor processing units that offer better energy efficiency. These technological advancements could help bridge digital innovation with climate imperatives in the coming years.
Source: techno-science.net