AI investment has not ended; wealth is simply being redistributed.

AI investment is not over; wealth is simply being redistributed. “The railway bubble also went through 30% to 50% corrections and surged three times.” 1. AI demand is growing, so why are semiconductor stock prices falling? It’s not because demand has peaked. The market is now recalculating not the growth rate, but the capital costs required to sustain growth and the ultimate scale of profits left for companies. 2. Tesla was the first to demonstrate this. Even if costs fall by 20%, if prices drop by 30%, profitability actually worsens. Higher productivity does not necessarily equate to shareholder benefits. 3. Memory chips may follow the same path. The supercycle doesn’t end only because of reduced demand; it ends when supply growth outpaces demand. 4. The same applies to AI tokens. Even if usage grows tenfold, if prices fall by 90%, revenue will stagnate. The key issue is whether usage grows faster or prices fall faster. 5. Meanwhile, high interest rates are compounding the pressure. When the yield on the 10-year US Treasury exceeds 5% and real rates remain high, adding a four-percentage-point credit spread, borrowing rates for AI infrastructure companies can reach 9%. A data center project with an 8% return on investment is profitable when financing costs are 4%, but with costs at 9%, further growth actually destroys value. 6. Securing GPUs and completing sellable computing power services for clients are two different things. According to someone from an AI lab, even if a top-level model isn’t produced, early secured GPUs have seen rents rise two to three times, and reselling them can still be profitable. 8. In summary, the AI industry is being squeezed from both directions. Technological advances and competition are pushing down product and service prices, while high real interest rates are keeping capital costs high. This results in a structure where prices fall but construction costs remain high. Meta has increased its capital expenditures, while Tesla has reduced theirs; today’s standard is no longer the scale of investment, but rather how much profit and how quickly one can recoup the invested capital. 9. The AI revolution will continue, but the companies leading the revolution may not be the same ones that achieve the highest stock returns. Just as in the internet era, value may shift from the companies building the infrastructure to those utilizing it. The current correction is likely not the end of AI’s growth, but rather a rearrangement of where the wealth created by AI will flow.#算力