What To Know
- The energy demands of AI are so high that they could potentially lead to an energy shortage.
- In North America alone, the electricity consumption of data centers increased from 2,688 megawatts at the end of 2022 to 5,341 megawatts by the end of 2023.
- This dramatic increase in energy demand forces power suppliers to rely on more polluting sources during peaks in usage, such as diesel generators.
The rapid rise of artificial intelligence (AI) is sparking concerns over its societal impacts and security risks. Yet, a new study sheds light on an often overlooked aspect: its environmental footprint.
the explosive growth of AI technology
Artificial intelligence, especially generative AI, is making groundbreaking changes in various sectors. From smartphones to home appliances, AI is being integrated into every conceivable device. However, this revolutionary technology comes with a significant yet less understood impact on the environment.
Researchers have delved into the environmental cost of AI, uncovering some startling facts. The energy demands of AI are so high that they could potentially lead to an energy shortage. The key culprit here is the massive computational power required by AI models, which have billions of parameters.
energy consumption: a ticking time bomb?
The energy consumption related to AI technologies has been a growing concern. Data centers, vital for running AI operations, have seen their energy usage skyrocket. In North America alone, the electricity consumption of data centers increased from 2,688 megawatts at the end of 2022 to 5,341 megawatts by the end of 2023.
- Globally, data centers consumed about 460 terawatts in 2022.
- This figure matches France‘s total electricity consumption and is expected to more than double by 2026.
This dramatic increase in energy demand forces power suppliers to rely on more polluting sources during peaks in usage, such as diesel generators.
the hidden costs: water and pollution
Beyond electricity, data centers also require substantial amounts of water for cooling purposes—about two gallons per kilowatt-hour of electricity used. This level of water consumption can significantly impact local ecosystems.
The environmental cost doesn’t stop there. The production and transportation of hardware used in data centers contribute heavily to pollution. Companies like Nvidia and AMD sold approximately 3.85 million graphics cards (GPUs) to data centers in 2023 compared to 2.67 million in 2022. These GPUs are essential for parallel processing tasks in AI but are complex and carbon-intensive to manufacture compared to standard CPUs.
a call for deeper understanding
The full environmental repercussions of AI remain difficult to assess entirely. Experts emphasize the need for more contextual methods to understand these impacts comprehensively. Due to rapid technological advancements, there hasn’t been sufficient time or opportunity to fully gauge or mitigate these effects.
