Google's AI search summaries use 10x more energy than just doing a normal Google search

Here is the link to the whole article from Boing Net

In recent years, large-language modeling algorithms have surged in popularity, with companies like Alphabet leading the charge. While these generative AI technologies promise innovative advancements, their environmental impact raises significant concerns.

The Hidden Energy Costs of AI Queries

According to Alex de Vries, founder of Digiconomist, AI-driven search queries consume approximately three watt-hours of electricity per query—ten times the energy usage of traditional searches. This power consumption is roughly equivalent to using a home phone for an hour. If AI-generated answers were added to all Google searches, the energy consumption could rival Ireland's.

Water Usage and Data Centers

Beyond electricity, the cooling requirements for data centers housing these AI models demand substantial water resources. Many of these facilities are in arid regions with inexpensive land, exacerbating the environmental strain. Reports suggest that many servers draw water from already stressed watersheds. One study estimates that global water demand for data centers could soon be half of the UK's total water usage.

A Broader Context of Environmental Impact

This situation mirrors the environmental concerns associated with other digital trends like cryptocurrency and NFTs. The push for less carbon-intensive energy sources is critical as global temperatures continue to rise. Yet, the resource demands of AI technology pose additional challenges.

Conclusion

The drive for AI innovation must be balanced with environmental responsibility. As companies continue to develop and deploy large-language models, it is imperative to consider and mitigate their ecological footprint. At Carbon ScaleDown, we are committed to providing solutions that help track and optimize the carbon footprint of AI applications, ensuring a sustainable future for technology advancements.