In this innovation, we reached this place. Graphics processing unit (GPU) that powers computing behind AI Lack of 99 % in cost Since 2006. With the wild estimates to increase electricity demand, there was a similar concern about the energy use of data centers in the early 2010’s. But not only the benefit of computing power and energy saving These estimates proved incorrectly But with a minimum increase in energy consumption, global computing capacity is worth 550 % increase from 2010 to 2018.
However, in the late 2010, the trends that saved us began to break down. As the accuracy of AI models is dramatic Better, Electricity required for data centers also started to grow rapidly. Now they are 4.4 % of the total demand, which is more than 1.9 % in 2018. Data centers use more than 10 % of the power supply in six US states. In Virginia, which has emerged as the center of data center activity, this figure is 25 %.
Estimates about future demands for power energy are uncertain and in limits, but in one study, Lawrence Berkeley National Laboratory estimates that data centers By 2028 US can represent 6 % to 12 % of total use of electricity. Communities and companies will see this type of rapid growth in electricity demand. This will put pressure on energy prices and the ecosystem. As a result of the estimation, many new gambling power plants have been called to develop or bring old people from retirement. In many parts of the United States, this demand will likely increase natural gas -powered plants.
This is a difficult situation. Still, when we zoom out, the use of AI expected electricity is still very small. America created about it 4,300 billion kW hours last year. In the next decade we will need another 1,000 billion to 1,200 billion or more. Almost Half of electricity additional demand Will be from electric vehicles. Another 30 % of the buildings and industry are expected to be from power -powered technologies. In the past decade, innovation in the vehicle and building has also increased, and this change will be good news for climate, communities and energy costs.
The remaining 22 % of the new power demand is estimated to come from AI and data centers. Although it represents a small piece of pie, it is most important. Because of their rapid growth and Geographic concentrationData Centers is the challenge of electricity that we now have to face – before we are able to do big things like cars and buildings before we have to know.
We also need to understand that we are buying energy consumption and carbon emissions associated with AI. Although the effects of developing semiconditters and strengthening AI data centers are important, they are much less than they are compared to Positive or negative effects AI can have petitions like electric grid, transportation system, buildings and factories, or consumer behavior. Companies can use AI to produce new materials or batteries that can better connect renewable energy into the grid. But they can also use AI to make it easier to find more fossil fuel. Claims about the potential benefits for climate are interesting, but they need permanent confirmation and will need help to understand it.
This is not the first time we have encountered the challenges that are competing Increase the demand for electricity. In the 1960s, US electricity demand was increasing by more than 7 % every year. In the 1970s, the growth was about 5 %, and in the 1980s and 1990s it was more than 2 % each year. After that, starting in 2005, we mainly used to develop flat electricity. In most of the next decade, our expected growth in electricity demand is 2 % again – but this time we have to do different things differently.
We need a “grid new deal” to handle these new energy demands, with enough capacity and intelligence to enhance the power system for the AI ​​to take advantage of public and private investments. New clean energy supply, investment in transmission and distribution, and strategy for virtual demand management can increase the emission, low prices and flexibility. Data centers that upgrade clean electricity and distribution systems can be fastened to connect the grid. Infrastructure banks can fund new transmission lines or pay to upgrade the current. Direct investment or tax concessions can encourage open data transparency about the standards of clean computing, the development of manpower in the clean energy sector, and data center operators about their energy use so that communities can understand and measure the effects.