AI’s insatiable hunger for energy is expected to soar in the coming years, potentially leading to power shortages.
A Gartner recent meaning last year, new servers required 195 terawatt hours of electricity. This is as much as 18 million households use in a year. But by 2027, the new servers will require 500 terawatt hours, or the electricity of 46 million households. The inevitable consequence of the impending energy shortage is the increase in the price of energy, which, according to Gartner, will also increase the operating costs of LLMs.
“IT giants are planning new, larger data centers to handle the massive amounts of data needed to train and infer the rapidly expanding large-scale language models (LLM) that underpin AI applications,” said Bob Johnson, an analyst at the company. will persist as new power transmission, distribution and generation capacity can take years to come on stream and future plans will not alleviate current problems. Major electricity users, in cooperation with large producers, strive to provide long-term guaranteed electricity sources independent of other network needs. In the meantime, the cost of electricity to run data centers will rise significantly as operators use economic leverage to secure the electricity they need. These costs will also be passed on to the providers of MI products and services.”
All of this has a negative impact on zero-carbon sustainability goals, because growing demand forces service providers to increase production by any means possible. In some cases, this means that fossil fuel power plants that are slated for retirement must be kept in operation even after their planned shutdown. “Increased CO2 emissions are expected to be generated in the short term to generate the energy needed for proliferating data centers,” Johnson said.
Data centers need a constant, stable power supply, which renewable energy sources – such as wind or solar energy – cannot provide. Reliable 24/7 power generation is only possible with hydroelectric, fossil fuel or nuclear power plants. In the long term, new technologies for better battery storage (e.g. sodium ion batteries) or clean energy (e.g. small nuclear reactors) can be relied upon to help achieve sustainability goals.
Source: sg.hu
