The rapid development of AI technology is making the US grid face unprecedented challenges, while also bringing new investment themes.

Due to the “poor profits” of Berkshire Hathaway Energy (BHE), a company that generates up to 30 gigawatts, Buffett said in a letter to shareholders this year: The future demand for the US grid and corresponding capital expenditure will be amazing. The opinion then drew widespread attention to the stability of the US grid and future electricity demand.

Buffett's concerns did not come out of nowhere. In recent years, concerns about America's “power grid crisis” have continued to spread. According to the US Electricity Forecast, the forecast for the growth of electricity demand in the US over the next five years is double that of a year ago.

AI data centers will become the main power consumers

Some media previously commented that the US is facing an imminent power grid crisis, which may be more urgent than the climate crisis. The article points out that new AI data centers, factory production under federal government subsidies, and the government-driven transformation of electric vehicles will be the main reasons for the surge in electricity demand.

AI data centers, in particular, are expected to put unprecedented pressure on the US power grid.

Currently, renewable energy sources such as wind energy and solar energy cannot provide this stability, and it takes 10 to 12 years to build transmission lines connecting remote renewable energy to the power grid. Therefore, electricity will be the main energy provider for data centers in the short to medium term.

Furthermore, data centers that support the AI training stage can consume up to 80 kilowatts, and data centers that support the AI inference stage consume half of the power, and data centers without AI requirements are only one-eighth of them — therefore, in the absence of more efficient training and inference processes, further development of AI will lead to increased data center consumption and increased power demand.

The next AI deal — energy

In this context, investors and analysts are looking for new investment opportunities related to AI.

Goldman Sachs analyst Louis Miller pointed out in the latest research report that as AI infrastructure construction progresses, the power infrastructure and energy sector will become a major trading topic:

“The focus of the entire artificial intelligence stack has moved from software and AI tools to hardware. Among them, the “winner-take-all” state is likely to decrease, and many existing companies in the market have the potential to increase profits through AI. As a result, we expect hardware and power-oriented industries to catch up with the broader topic of artificial intelligence.”

Specifically, Miller believes it is possible to focus on four major areas: smart grid infrastructure (30% weight), unregulated utilities (weight 25%), raw materials for power generation (weight 25%), and other utilities (weight 20%). Companies in these fields may receive new orders and revenue streams as demand for electricity increases.

Furthermore, the report said that as electricity demand grows, demand for power generation raw materials such as copper and uranium is also expected to surge.

According to reports, these materials are essential for electricity transmission and the construction of new nuclear power plants. As a result, producers of related materials may also become new favorites for investors.

The report concludes:

In the short term, regardless of whether software companies successfully launch their AI products, hardware and power companies will win this race.

Editor/Somer