[Energy-topia]AI-Driven Power Surge: Are We Prepared?

At the World Economic Forum (WEF) annual meeting held in Davos, Switzerland last January, Sam Altman, CEO of OpenAI, stated, "Artificial intelligence (AI) consumes much more energy than people think," emphasizing the need for an energy breakthrough for the future of AI.

He added that there must be something on a massive scale, such as nuclear fusion or radically cheap solar power, that no one has yet planned. However, he did not specify how much energy AI would require.

Soon after, the International Energy Agency (IEA) provided an answer in its 'Electricity 2024' report. The report projected that the global data center electricity demand, which was 460 TWh (terawatt-hours) in 2022, would expand to between 620 and 1050 TWh by 2026. Depending on the scenario, this means it could more than double in just two years.

460 TWh is roughly equivalent to the annual electricity consumption of France. If it expands to 1000 TWh in two years, it would be comparable to Japan's annual electricity consumption.

The IEA identified AI and cryptocurrencies as the causes of the explosive electricity demand from data centers. Enormous-scale data centers are required to provide AI services.

Analysis data from the International Energy Agency (IEA) predicting a sharp increase in power demand for data centers due to artificial intelligence (AI) and virtual currencies. By 2026, it is expected to increase to 620~1050 TWh depending on the scenario. Image source=IEA

In particular, the expansion of generative AI like OpenAI's ChatGPT is expected to further drive up electricity demand. According to research, a single Google search requires 0.3 Wh of electricity. In contrast, each query to ChatGPT consumes 2.9 Wh of electricity. This means ChatGPT uses about 10 times more electricity than a Google search. Based on this, the IEA estimated that if Google searches were replaced by AI like ChatGPT, an additional 10 TWh of electricity would be needed annually.

Big tech companies have begun preparing for the AI-driven surge in electricity demand. According to an analysis by the British weekly magazine The Economist, Alphabet (Google's parent company), Amazon, and Microsoft (MS) invested $40 billion in the first quarter of this year, with most of the investment focused on data centers to handle AI. These companies are competitively investing in the energy sector to supply electricity to their data centers. They do not discriminate among types of 'clean energy' such as solar, wind, nuclear, or geothermal.

On the 1st of this month, MS and Brookfield agreed to invest in 10.5 GW of renewable energy in the U.S. and Europe by 2030. MS also signed a contract with Constellation Energy, the largest nuclear power operator in North America, to supply electricity to data centers when solar, wind, and other renewable energies are unavailable.

Altman has invested $375 million in Helion, a nuclear fusion startup, and is also pushing for the listing of Oklo, a small modular reactor (SMR) company. National-level efforts are also underway. Sweden, which owns 60 data centers, announced plans last August to build data centers powered by SMRs.

The Ministry of Trade, Industry and Energy plans to announce the 11th Basic Plan for Electricity Supply and Demand (Electricity Plan) later this month. Due to political sensitivity, the announcement was postponed until after the general election. South Korea has been debating for years between renewable energy and nuclear power. Compared to other countries that are mobilizing all possible means to prepare for the AI-driven electricity surge, the situation here seems remarkably complacent.

This content was produced with the assistance of AI translation services.