Key Takeaways:

1. A gradual shift. After a rapid surge in the first half of the year, global asset prices tied to AI have recently seen a notable correction, impacting overall risk appetite in financial markets. The question is whether this is merely a temporary technical adjustment or a fundamental re-evaluation of AI investment's core nature. As AI's development logic appears to increasingly mimic human learning, what can humanity learn from AI's own evolution?

2. Different economics from the internet era. During the internet revolution, once the "information superhighway" was built, marginal costs decreased or even hit zero. In contrast, AI relies on concentrated computing power to achieve breakthroughs, and even after infrastructure and model training are complete, its applications continue to consume significant resources, not to mention the phase of increasing capital expenditure. AI's marginal costs may be rising, with a clear profile of "heavy assets, high debt, and slow returns."

3. Financial pressure points. While the current profits of major global AI companies are acceptable, the immense demand for capital expenditure is weakening their free cash flow. If this is combined with rising external financing costs, it could increase market concerns about their ability to deliver financial results. Unlike the internet era, the AI supply chain is a cross-border network of multiple oligopolistic markets, where any intermediate link can trigger overall volatility.

4. Looking ahead. The long-term trend in technology remains intact, but investment is not limited to AI. From a macroeconomic perspective, investment is a function of interest rates. While China's financing costs are steadily and slightly declining, the change in corporate investment returns appears to be steeper. Under the demand for investment to "stabilize and recover," interest rates and returns will chase each other and eventually converge. Particularly during a period of capital market support, funding rates seem to have a characteristic of being easier to loosen than to tighten.

Full Article:

After a rapid surge in the first half of the year, global asset prices tied to AI have recently seen a notable correction, impacting overall risk appetite in financial markets. The question is whether this is merely a temporary technical adjustment or a fundamental re-evaluation of AI investment's core nature. As AI's development logic appears to increasingly mimic human learning, what can humanity learn from AI's own evolution?

What Kind of Investment is AI?

During the internet revolution, once the "information superhighway" was built, marginal costs decreased or even hit zero. In contrast, AI relies on concentrated computing power to achieve breakthroughs, and even after infrastructure and model training are complete, its applications continue to consume significant resources, not to mention the phase of increasing capital expenditure. AI's marginal costs may be rising, with a clear profile of "heavy assets, high debt, and slow returns."

While the current profits of major global AI companies are acceptable, the increasing demand for capital expenditure is weakening their free cash flow. If this is combined with rising external financing costs, it could increase market concerns about their ability to deliver financial results. Unlike the internet era, the AI supply chain is a cross-border network of multiple oligopolistic markets, where any intermediate link can trigger overall volatility.

Is Investment Only About AI?

The long-term trend in technology remains intact, but investment is not limited to AI. From a macroeconomic perspective, investment is a function of interest rates. While China's financing costs are steadily and slightly declining, the change in corporate investment returns appears to be steeper.

Looking ahead, the asset allocation logic between AI and non-AI sectors may be reshaped. Under the demand for macro investment to "stabilize and recover," domestic interest rates and corporate investment returns will chase each other and eventually converge. Particularly during a period of capital market support, funding rates seem to have a characteristic of being easier to loosen than to tighten.