Artificial intelligence can consume more energy than bitcoin mining – computers
According to researcher Alex de Wreaks-Gao, a doctorate candidate from the Amsterdam Environment Studies Institute, Artificial Intelligence is soon cryptocurrency mining, which is the most famous bitcoin. The researcher, the organization founded by the Digiconomist has been dedicated to reveal the unintended consequences of developing technologies.
In the study PublishedThe researcher focuses on the consequences of the stability of technologies such as blockchain and artificial intelligence. The Analysis concluded that by the end of 2025, half of the electricity spent for global data centers could be consumed by the end of 2025.
Investigating some of the giants of AI technology, such as Google, who has faced crises regarding fuel capacity. Both Google Microsoft reports the increase in power consumption and emissions as submitted to their 2024 account, pointing out how the growth of artificial intelligence is.
But in relation to concrete data, companies do not look completely transparent. Most of the time reports usually refer to the widespread increase in power consumption of data centers, not specifying their use. Jim is asking for good data to understand what real energy costs with AI are.
The researcher claims that fuel consumption requirements for AI have been duplicated for the year. AI acceleration modules can reach 23 GW until the end of 2025, which exceeds bitcoin mining, which represents half of the total electricity spent in data centersWith the exception of crypto mining. This is due to the low demand of mining through the current hardware, although most sector companies do not want to abandon previously invested parts to become more effective.
In an interview The edgeWere, and,, and,, and,, and,.. Alex de Wreaks-gao companies indicate that the idea of ”big is good” and therefore they are constantly increasing the size of their models, to be better in the market, but at the same time creates more resources needs. It has succeeded in raising data centers, especially in the United States, new types of power, including nuclear reactors, already speaks.
Talk about the power required for AI data centers is not from now. More demanding in terms of consumption of artificial intelligence, manufacturing The estimate of the construction of nuclear centers to nourish this technology. In early February, a Idtechex search Indicated that the continuous growth of artificial intelligence would be the result of More than 2,000 TWH power consumption by data centers in 2035Anything like 2 million GWH. Add this to the growth of CO2 emissions as a result of the feeding of these new data centers.
AS The good news is that the latest progress of AI models also begins even more effectively. The Deepseek has already stated that its model of Lama 3.1 has been used by its model target. And also A more effective distillation version has been announced today.
