Many people are focused on AI models and chip hardware competitions, but they overlook the most fundamental question—what do these computing devices consume?

The answer is simple: electricity.

Whether it's training large models or inference services, whether it's data centers or cloud computing infrastructure, ultimately all costs boil down to three words: how much is used. But more importantly, this demand is entering an exponential growth phase.

Just look at the current situation. Major countries and regions around the world are rapidly expanding their computing power centers, which directly drives up electricity demand. This is not a short-term trend for this year or next year, but a structural trend for the next 5-10 years. Electricity consumption will rise along a relatively stable but continuously increasing slope—that's certain.

The market will ultimately hype up the most fundamental and hardest-to-replace link. While everyone is still debating which model is stronger, true capital has long been systematically focusing on the energy line—generation capacity, transmission and distribution infrastructure, energy storage solutions, grid upgrades and renovations. These will become key areas for strategic deployment.

This means that the AI market may not only stay within the rotation of the tech sector but could also evolve into a systemic reevaluation of the entire energy industry.