By Şeref Kılıçlı
As the development of artificial intelligence applications increases the electricity consumption of data centers, global technology companies are looking to nuclear power as a solution. However, Amazon’s exclusive agreement to purchase electricity from a nuclear power plant was not approved, which will affect the methods.
Although the first thing that comes to mind when we think of nuclear technology is ‘deterrent weapons’ due to the influence of world politics, it also has a wide range of uses such as health, agriculture, environmental pollution and energy for peaceful purposes. Within the scope of nuclear technology, studies are being carried out in many areas such as the fight against cancer, the development of X-ray devices, the protection of water resources with isotopic techniques, the application of drip systems in regions with water shortages, the measurement of pollution in the seas and the determination of its causes. Nuclear energy is considered by many countries as an affordable, reliable and clean energy source that can help mitigate the impact of the climate crisis. The fact that nuclear accidents, which have marked the last 80 years of the world’s history, have caused serious long-term damage emphasizes the need for greater safety and security in new technologies.
NUCLEAR PRODUCTION
Last year, 2600 terawatt-hours of electricity generated from nuclear power met 9 percent of global demand. The International Energy Agency (IEA) expects global nuclear power generation to exceed the previous record set in 2021 by 2025. Although some countries are phasing out nuclear power or retiring plants early, nuclear output is expected to grow by an average of close to 3 percent a year through 2026, driven by the start of commercial operations of new reactors in several markets, including China, India, Korea and Europe. It is noted that the completion of maintenance work in France and the restart of nuclear production at several plants in Japan will also affect the increase. Many countries are making nuclear energy a critical part of their energy strategies as they aim to reduce greenhouse gas emissions and maintain energy security. Indeed, at last year’s 28th Conference of the Parties (COP28) to the United Nations Framework Convention on Climate Change, 22 participating countries pledged to triple their nuclear power capacity by 2050 to achieve net zero emissions.
DATA CENTERS
According to the IEA’s Electricity 2024 report, global electricity demand is expected to increase at a faster rate over the next three years, growing at an average annual rate of 3.4 percent until 2026. According to the report, electricity consumption from data centers, artificial intelligence (AI) and the cryptocurrency sector could double by 2026. While data centers are important drivers of growth in electricity demand in many regions, their total electricity consumption globally could be over 1,000 terawatt-hours (TWh) in 2026. For comparison, this demand is equivalent to Japan’s electricity consumption.
2 PERCENT OF CONSUMPTION
The widespread use of artificial intelligence leads to an increase in energy consumption. The primary reason for this increase in energy consumption is the energy-intensive nature of data centers, which are essential for AI operations. According to another report published by Goldman Sachs, data centers worldwide currently consume about 2 percent of total energy. However, this rate is expected to rise to 3-4 percent by 2030. Again, according to Goldman Sachs’ research, the energy needs of data centers operating in the US will increase by 160 percent by 2030, accounting for 8 percent of the total energy produced by the US. While attention is drawn to the increase in electricity consumption from artificial intelligence data centers, technology companies are looking for the solution in nuclear, which can provide uninterrupted power. For this reason, technology giants, including companies such as Google, Microsoft and Amazon, have turned their attention to nuclear energy investments.
SURPRISE DECISION IN THE USA
In March, Amazon, seeking a solution to its growing energy needs, paid $650 million for a data center adjacent to the Susquehanna nuclear power plant owned by Talen Energy and signed a long-term agreement to purchase power from the plant. But on November 1, the US Federal Energy Regulatory Commission (FERC) rejected the exclusive deal that would have allowed the data center to use more power from the adjacent nuclear plant. Commissioners said that the modification plan submitted by the regional grid operator on behalf of the parties did not sufficiently demonstrate why the special agreement should be allowed under federal rules. The decision is seen as a blow to big tech companies’ efforts to power their data centers with electricity from power plants located next to their facilities. Amazon also recently announced that it had signed three new agreements with different energy companies involving small modular reactors (SMRs) to support the development of nuclear energy projects and meet its needs.
DATA CENTER CONSUMPTION IN EUROPE COULD TRIPLE
According to a report published by McKinsey in October, the energy consumption of data centers in Europe is expected to nearly triple by 2030. This is expected to lead to a major increase in the supply of electricity from low-carbon sources. In the next 6 years, data centers are expected to account for about 5 percent of total energy consumption in Europe, while today this rate is around 2 percent. The report also emphasized that at least 250-300 billion dollars of data center infrastructure investment, excluding electricity generation capacity, is required to meet the energy demand of data centers.
HOW MUCH ELECTRICITY DOES ARTIFICIAL INTELLIGENCE CONSUME?
According to data released by Google, the total energy consumed for every 100 searches on Google is equal to the energy consumed by a 60-watt light bulb in 28 minutes. According to data from the International Energy Agency (IEA), queries made with artificial intelligence applications consume much more energy. Accordingly, a ChatGPT query consumes 2.9 watt hours of electricity, while a Google search consumes 0.3 watt hours. This means that AI queries consume about 10 times more energy than Google searches. In other words, the energy consumed by a 60-watt light bulb in 28 minutes can be accessed in 10 artificial intelligence processes.
SMR INVESTMENTS ARE ALSO A PRIORITY
Google, which also prioritizes nuclear investments, announced that it will use small modular reactors (SMR) to meet the energy needs of its artificial intelligence data centers. The company signed an agreement with a company called Kairos Power for this purpose. According to the agreement, the first reactor is expected to be operational by 2030 and the others by 2035. The total number of SMRs will be 8. Microsoft signed an important agreement with Constellation Energy to reopen the Three Mile Island Nuclear Power Plant, which was closed in 2019 to provide the energy needed by artificial intelligence. Microsoft also announced that it will purchase all 835 megawatts of power from the plant for the next 20 years. The plant is expected to be restarted by 2028 with a $1.6 billion investment by Constellation Energy. Oracle is designing a data center that will require more than 1 gigawatt of electricity and has received permission to build 3 SMRs to power this data center. OpenAI, the owner of ChatGPT, is investing in nuclear fusion.
Source: https://istanbulticaretgazetesi.com/yapay-zeka-enerji-icin-nukleere-yoneldi-