The biggest problem for people about AI is energy consumption, and a recent report says that matters are only worsening in this regard in the next few years. However, the same report also argues that there is a silver layer.
A study by the Internal Energy Agency (IEA) titled Energy and AI was recently published (Via Benefactor) The use of global datases and consultation data with “governments and regulators, tech sector, energy industry and international experts”.
In this, the paper shows that energy demands for data centers will increase largely by double, and in particular, the use of AI data centers will increase by the element of four.
The energy needed to provide these data centers has increased from 460 Twh to 2030 in 2024 more than 1000 Twh. It is then predicted to reach 1,300 Twh by 2035.
Although traditional data centers are likely to grow over time, AI-UPTMized servers are taking the biggest part in this growth.
As indicated by the IEA Director Dusty, Laura Kozi, Full presentation On paper, the energy demand of individual data centers is increasing over time.
Hyper Scale Data Centers (the largest) use equal power of 100,000 households today, which is mostly ready for equal use of 2 million households. Currently the largest (but not yet under construction) will use the energy of 5 million households.
However, the article argues that it is not all the punishment and depression in relation to the change of climate. It starts’AI and climate change‘The section is stating that more than 100 countries have deals to reach the pure zero emissions between 2030 and 2070.
When the demand is now increasing, forty years distance is a minor, and ‘Net Zero’ can mean a lot, massively.
The fact is that bringing energy requirements does not appear here, it shows that contrary to the rule on AI, it will be achieved with pure zero offsets.
However, the argument made in ‘Energy and AI’ is the most part of the AI model, and their development, can be used to restrain it in other areas of energy, which can reduce emissions in terms of methane, power sector and large industry.
The report argues that “adoption of existing AI applications in the closing use sectors can lead to a reduction in 1,400 MT CO2 emissions in 2035”. The report estimates global fuel combustion emissions in 2024.
The offer of this argument includes a adoption, which the report shows that it is not currently happening. “It is important to note that there is currently no speed that ensures these AI applications widely adopted. Therefore, their overall effect, even in 2035, can be modest if the necessary conditions are not created.”
Talking to The Guardian, Claude Terms, former secretary of Luxembourg’s sustainable development and infrastructure, criticizes the results of this report.
“Instead of giving practical recommendations to governments, how to organize and reduce the major negative effects of AI and new mega data centers on the energy system, the IEA and its (Chief) winner Birul are giving a welcome gift to the new Trump administration and the patronizing tech companies patronizing this new US government.”
The IEA report has a built -in AI -Chat Boat on Page and when I asked what the paper says about climate change, he told me:
“Widespread adoption of existing AI applications can lead to a reduction in emissions, which can reduce the equivalent of 5 % of energy -related emissions in 2035. However, it is much smaller than it is needed to cope with climate change.”
