I. Goldman Sachs' K-shaped Divergence Theory: Identifying the Right Problem, but Overestimating the Role of AI
Goldman Sachs reports that the Chinese economy is currently experiencing a situation where "technology is booming while domestic demand is weak"—tech exports are increasing, but real estate and traditional consumption are sluggish. This observation is accurate, but the report atributes the weakness in consumption to AI, which contains several logical flaws:
- AI is not the sole reason for the decline in manufacturing employment: Although manufacturing has lost 20 million jobs over the past decade, the working-age population has decreased by 54 million (more than the number of jobs lost), and there has also been a shift in the supply chain. On the other hand, the information technology services sector has added 5.1 million jobs, and AI itself has created new roles such as data annotators.
- The decline in the real estate market cannot be blamed on AI: It is due to the collapse of developers and the suspension of construction projects, which have led to fewer jobs in the construction industry, not because AI has taken over the work of construction workers.
- Weak consumption is due to a lack of willingness to spend, not a lack of funds: Household savings have increased from 93 trillion yuan in 2020 to 140 trillion yuan by mid-2024. People are just cautious about spending because they have poor expectations, not because their balance sheets have collapsed (if there were a real collapse, they wouldn't be saving money).
In short, Goldman Sachs has identified the phenomenon of "booming technology and weak domestic demand," but it has misinterpreted the causal chain, blaming AI for something it shouldn't.
II. The Two Major Constraints on the Current Economy: Old Problems Unresolved, New Contradictions Emerging
Although there are issues with Goldman Sachs' reasoning, the risks associated with a K-shaped divergence do need to be addressed. There are two inescapable constraints on the economy:
1. Old Problem: Declining Investment Efficiency
There is an indicator called ICOR (the amount of capital required to generate one unit of GDP). Twenty years ago, it took 3.4 yuan to generate one yuan of GDP; now, it takes 8.3 yuan—more money is being invested with less output. In contrast, the United States and South Korea have seen improvements in investment efficiency over the long term (around 3-4 yuan per unit of GDP). China's investment efficiency has been declining over the past decade, indicating that the old approach of relying on large-scale infrastructure projects is no longer effective.
2. New Contradiction: The AI Efficiency Paradox
While AI can help companies save labor and improve efficiency, people are the main drivers of consumption. If AI replaces workers, those workers will lose their jobs, and those still employed may fear salary cuts, which will deter them from spending. As a result, companies produce more and cheaper goods, but no one buys them. The more technology advances, the harder it is for companies to sell their products, and the AI industry could end up being a bubble.
Worse still, these two problems can create a vicious cycle: AI replaces workers → consumption declines → companies earn less → they are reluctant to invest → GDP falls → they are forced to rely on traditional investments → ICOR worsens.
III. Breaking Free from the "Either/Or" Dilemma: The Third Option is to "Invest in People"
The academic debate about whether to focus on large-scale infrastructure or stimulate consumption leads to dead ends. The real solution is not to stop investing altogether but to shift the direction of investment—investing in people so they feel confident enough to spend. This is the type of investment that offers high returns:
- It's not about distributing money indiscriminately but targeting areas that boost living standards and purchasing power, such as providing subsidies to low-income groups and reducing taxes. These funds will directly contribute to consumption, breaking the vicious cycle of "unemployment → lack of spending → company bankruptcies → more unemployment."
- By maintaining a large market of 1.4 billion people, there will be a demand for new AI products, providing sustained momentum for technological advancement.
IV. A Three-Step Policy Approach: From Stabilizing the Economy to Creating a Positive Cycle
The author proposes a three-step strategy aimed at promoting "dual-driven growth through consumption and innovation":
1. Short-term Emergency Measures:
- Provide monthly subsidies of 500 yuan to low-income and extremely poor individuals, costing over 260 billion yuan per year (only 1.2% of fiscal revenue) to maintain basic consumer spending.
- Raise the personal income tax threshold to 8,000 yuan, which could unleash an additional 500 billion yuan in potential consumption.
- Allow small and medium-sized enterprises to hire AI-replaced workers for half a year without paying social security fees to prevent the spread of unemployment.
2. Medium-term Rebuilding:
Transfer 40% of the consumption tax revenue to local governments, giving them incentives to stimulate consumption—after all, it directly affects their finances and is more effective than administrative orders.
3. Long-term Sustainable Growth:
China's large market of 1.4 billion people is a significant advantage. If a new AI product succeeds in the domestic market, the scale effect can reduce research and development costs, allowing companies to compete globally with better value for money. By first stabilizing domestic consumption, companies will have the profits needed to develop the next generation of AI technologies, creating a positive cycle of "consumption → corporate profitability → further innovation → more consumption."
Conclusion
The core message of this article is that AI should not be scapegoated for the current economic issues. The real problems lie in the inefficiency of the old investment models and the failure to convert technological advancements into consumer spending. The solution is not a simple "either/or" choice but to invest in people, thereby promoting a mutually reinforcing relationship between technology and domestic demand and breaking out of the K-shaped divergence.