The current AI market trend is a very typical three-stage process:

Stage One: Computing power and hardware win first. This includes Nvidia, HBM, memory, storage, optical modules, and the server supply chain.

Stage Two: Diffusion to power and physical infrastructure. This includes data centers, cooling, power supplies, transformers, power grids, nuclear power, natural gas, land, and grid connection resources.

Stage Three: Applications begin to differentiate and emerge as winners. It's not "all AI applications," but those companies that can convert AI into revenue, profit, regulatory approval, clinical success, or network effects.

Therefore, capital will rotate through these areas, but not mindlessly to all application stocks. More precisely: the infrastructure chain earns money from certainty and scarcity; the application chain will earn money from explosive revaluation after validation in the future. Right now, the market is still rewarding "who can collect money today," and the next stage will reward "who can turn AI into long-term monopoly cash flow."