Archive — JP — japantimes
We need energy for AI, and AI for energy
The rapid advancement of artificial intelligence (AI) necessitates vast amounts of energy, as a single AI query uses significantly more electricity than traditional web searches. Current AI infrastructure already consumes about 3% of U.S. electricity, with projections indicating this could exceed 6% in the next decade. If energy demands surpass supply, AI development may be hindered.
Key points
- AI requires substantial energy for its operation; a single ChatGPT query uses 10 times more electricity than a conventional web search.
- AI’s electricity consumption is projected to rise from 4 TWh in 2023 to 93 TWh by 2030, nearly equivalent to the total energy use of Washington State in 2022.
- Current AI data centers contribute around 3% of annual U.S. electricity use, with estimates suggesting this could double in the next 5-10 years.
- If energy demand from AI exceeds supply, it threatens the progression of AI development.
Why it matters
- Historically, technological advancements have depended on energy availability - for instance, the Industrial Revolution's reliance on coal.
- Current trends in IT, such as increased cloud computing and the rise of smart devices, further emphasize the growing need for energy-efficient technologies.
- By 2025, addressing the energy needs of AI could become a critical factor in global energy policies, similar to the transition towards renewable energies.
The interdependence of AI development and energy availability highlights the urgent need for a sustainable energy solution to ensure the continued progress of AI technologies in the coming years.