According to reports from the financial information app, Intelligent finance, technology giant Alphabet Inc.'s Google (GOOGL.US) and the utility company located in Nevada, USA, owned by the stock god Buffett, as well as a start-up focusing on geothermal energy, have reached important cooperation agreements to use efficient and clean energy geothermal power to drive Google's large data centers, which are mainly based on AI training/inference workloads. The booming development of data centers has greatly increased the electrical demand all over the United States.

According to a statement released by Google, energy start-up Fervo Energy will develop a new geothermal power plant to provide up to 115 megawatts of power to NV Energy, a subsidiary of Berkshire Hathaway Energy Company, owned by Warren Buffett. NV Energy will then sell the power resources to Google, but the multi-party agreement still requires formal approval by national regulators.

Unlike solar and wind energy infrastructure, geothermal power is not affected by weather and seasons, and can provide stable baseload power, so geothermal power plants can run 24/7. They mainly use water that is instantaneously heated by underground rocks for power generation, without producing a large amount of greenhouse gas emissions.

Geothermal energy is the natural release of thermal energy inside the Earth, which will last for billions of years throughout the Earth's lifetime. Therefore, from the time dimension of the Earth, geothermal energy is a renewable resource, and geothermal power generation in the specific operation process almost does not emit carbon dioxide and other greenhouse gases, making it a highly efficient and clean energy.

Geothermal energy comes from the superhot magma inside the Earth, which is transferred to the surface through cracks or hot springs in the crust. Energy start-up Fervo Energy uses horizontal drilling in geothermal reservoirs to increase power output.

Currently, large-scale data centers, led by technology giants such as Google, Microsoft, and Amazon, are consuming more and more electricity, which brings great pressure to technology companies that have set goals to reduce carbon emissions that cause global warming. Therefore, these technology companies have turned their attention to renewable resources that focus on clean attributes. Statistics show that the power capacity purchased by Google from NV Energy and Fervo can supply electricity to about 86,000 households.

In the era of AI, clean energy supply is becoming increasingly important.

By 2023, ChatGPT will be popular worldwide, Sora's video model will be launched in 2024, and NVIDIA, the 'shoveler' in the AI field, will achieve unparalleled performance for multiple quarters, which may mean that human society will gradually enter the AI era from 2024. Data centers play a vital role in the full-blown rise of generative AI such as ChatGPT. As the market demand for AI data centers accelerates, data centers, which were already 'power monsters', are expected to have an increased demand for energy. The International Energy Agency predicts that by 2026, the total electricity consumption of global data centers will increase from about 460 terawatt-hours in 2022 to at least 1,000 terawatt-hours.

A research report released earlier this week by Goldman Sachs showed that AI is expected to drive a crazy 160% increase in data center power demand by the end of 2030. Goldman Sachs predicts that US utility companies alone will need to invest about $50 billion in new generation capacity to support AI data centers.

"On average, the power supply required for AI services such as ChatGPT query and inference is almost 10 times that of normal Google search. In this difference, how the US, Europe, and the world consume electricity - and how much it will cost - will undergo earth-shaking changes," Goldman Sachs said.

"For many years, the demand for electricity at data centers has been very stable, even as their workloads have increased. Now, with the slow-down of the speed of improvement of the power usage efficiency and the rise of the AI revolution, we expect data center power demand to grow by 160% by 2030," Goldman Sachs added.

According to Boston Consulting Group's forecast data, the share of data centers in US electricity consumption is expected to double, from 126 terawatt-hours in 2022 to 390 terawatt-hours in 2030. Terawatt-hours are used to describe the largest level of electricity consumption, typically for national energy statistics, large-scale energy project planning and evaluation. Large industrial facilities, such as super steel mills, may not consume more than 10 terawatts of electricity per year.

The strong demand for clean energy from large data centers such as Google and Microsoft is mainly due to the global decarbonization trend, which focuses on renewable resources such as wind power and geothermal energy with clean attributes. According to the latest statistics from the energy think tank Ember, the expansion pace of renewable energy such as wind and solar energy continues to far exceed traditional fossil fuels such as coal, and the share of renewable energy in global electricity capacity reached a record-high 30% last year. The think tank predicts that this trend will continue to accelerate this year.