1. Notizie
  2. AI consumes huge amounts of Energy

AI consumes huge amounts of Energy | Microsoft Generative ai Course | Generative ai Tools Github | Bing ai | Turtles AI

AI consumes huge amounts of Energy
Editorial Team28 June 2024

 

 


THE AI AND THE ENERGY CRISIS : CHALLENGES AND SOLUTIONS 

 

The growing demand for electricity to power the computing clusters needed to handle large language models is a major issue for artificial intelligence specialists. Ami Badani, marketing director of Arm Holdings, warned that AI could consume a quarter of U.S. energy by 2030. According to the International Energy Agency, electricity consumption in the artificial intelligence, data center and cryptocurrency sectors could equal that of Japan by 2026.

During Fortune’s Brainstorm AI conference in London, Badani pointed out that ChatGPT consumes 15 times more energy than a normal web search, highlighting the need to address the energy problem in order to continue AI advances. As generative AI is applied to mobile devices, it will be critical to scale without overloading the power grid. As an example, Badani mentioned Sora, OpenAI’s generative video AI, which requires 100,000 AI chips at their maximum energy capacity just for training.

The technology industry is developing more and more powerful tools, increasing the demand for processing and thus energy consumption. Currently, AI data centers account for 2 percent of global electricity consumption. The increasing reliance on data centers could put pressure on existing energy infrastructure, causing potential electricity shortages.

Demand for data center capacity is already rising, with record levels in Europe in 2023 and a doubling of net absorption in North America. Daniel Thorpe, data center researcher at JLL, predicted that more data will be generated in the next five years than in the previous decade, increasing demand for storage capacity.

Ivo Ivanov, CEO of DE-CIX, predicted that the United States will need 50 percent more data center capacity by 2027 than in 2020. Existing infrastructure will need to be upgraded to support new AI technologies, with increasing use of liquid cooling to keep temperatures down.

Solutions to the energy problem include the development of semiconductor chips optimized to consume less power. Paradoxically, artificial intelligence itself could offer solutions to optimize the power supply chain and improve technology, reducing space, power and cooling requirements in data centers.