The growth of the data center largely stimulated by the generating AI should cause an increase in global energy demand in the coming years and a report from this week of the International Energy Agency provides new estimates How much.
The amount of electricity necessary for data centers around the world should more than double to around 945 terawatt hours in 2030, which is more than what the whole country of Japan consumes today, according to the report.
The popularity of the AI generation tools as an Openai chatgpt and Google Gemini has skyrocketed in recent years. These large models of language and their parents require a huge amount of calculation power, operating on high -end graphic processing units like those manufactured by Nvidia. These not only need a lot of electricity to work, but they also generate heat – which means that even more energy is necessary to keep them cool. All this adds up quickly.
In the United States, about half of increased electricity demand in the United States should be for data centers. Processing data should need more electricity than manufacturing all “energy intensity products” – aluminum, steel, cement and chemicals, said the IAI.
“AI is one of the largest stories in the world of energy today – but so far, decision -makers and markets have lacked tools to fully understand the impacts in vast,” said Fatih Birol, executive director of the IEA, in a press release.
AI energy consumption has worried the decision -makers
The energy demand for AI is well documented and in the minds of decision -makers and experts for years. This week, the Chamber’s Energy and Trade Committee held a audience On how to provide enough energy for data centers while members of the Congress are trying to understand the problem.
Experts have said that the American electricity system will require improvements and changes to meet growing demand after relatively stable decades of consumption. An area of interest in the Committee was to provide “basic power” – a constant electricity power supply throughout the day. This desire for coherence has aroused questions from members about the question of whether renewable energy sources such as solar and wind energy would work or if growing demand should be satisfied by new power plants burning fossil fuels such as natural gas. The combustion of these fossil fuels is a major engine of climate change.
Data centers like this in Ashburn, Virginia, are becoming more and more numerous as the demand for calculation power is developing to follow the new generative AI models.
There is also the question of whether the existing American electrical network can manage the necessary demand and supply.
“Even without the anticipation of rapidly increasing demand for rapid electricity, the American electrical network needs modernization investments,” said Melissa Lott, professor at Climate School at Columbia University. “Recent forecasts for increasing power demand strengthen these even more urgent and necessary investments.”
The calendar of the availability of electricity, given the overvoltages of demand sometimes like the heat waves and the daily life of the solar generation, is a problem that can be treated in various ways. Lott highlighted energy efficiency efforts, such as energy devices, and demand for demand reduction also known as virtual electric power plants. These can reduce energy demand and level peaks and valleys throughout the day.
Can AI solve its own energy problem?
The AIE report suggests that generative AI could, over time, help solve the problem caused by its energy request. If AI accelerates science and technological improvements, this could lead to better solar panels and batteries, for example. “The challenges of energy innovation are characterized by the types of problems that AI is good to solve,” said the IEA report.
But another solution could reside in how AI data centers use power. A Report earlier this year Researchers from the University of Duke have suggested that AI data centers can be more easily extinguished and on the way to adapt to the needs of the electrical system, allowing grid operators to adapt more easily to growth.
“This analysis should not be interpreted to suggest that the United States can fully respond to its electricity requests to near and medium term without strengthening the peak capacity or expanding the network,” wrote the researchers. “On the contrary, he stresses that flexible load strategies can help exploit the height of the existing head to integrate new charges more quickly, reduce the cost of expansion of capacity and allow you to focus more on the highest investments in the electrical energy system.”
