Summary of Key Points
This article compares two phenomena: the bursting of the new energy vehicle (NEV) bubble and the inflating AI bubble, revealing a common underlying logic: the technologies themselves are genuine and valuable, but capital investment far exceeds actual demand, ultimately leading to the burst or risk of a bubble. The NEV bubble has already had its consequences (bankruptcies of car companies, 850,000 car owners left without support, and financial losses for local governments). The AI bubble, on the other hand, is expanding at a larger scale with a higher degree of financialization, and if it bursts, the impact could be much broader, affecting global bond markets, private lending, semiconductor supply chains, and many other sectors. The core message is that technology can endure economic cycles, but capital cannot; when bubbles burst, the groups at the end of the chain bear the greatest cost.
I. The NEV Bubble: A Lesson from the Past
The NEV bubble has completely burst, with the most evident signs being severe overcapacity and a large number of car companies going bankrupt:
- Overcapacity: The total national automobile production capacity is 47.5 million units, but there is a supply-demand gap of over 13 million units, meaning that three out of every ten cars produced cannot be sold. Leading companies like BYD and Tesla can operate at full capacity, but the utilization rate for mid-tier companies is less than 50%, and for smaller ones, it's even lower than 30%.
- Massive bankruptcies: Of the 129 NEV brands, most have annual sales far below the break-even point of 400,000 units. Well-known brands such as NIO (the top seller in 2022), WM Motor, and HiPhi have gone bankrupt, leaving 850,000 car owners without support—car systems lose connectivity, batteries fail to be repaired, and 4S stores close, turning cars into scrap metal.
- Financial losses for local governments: Cities like Yichun and Tongxiang provided incentives such as building factories for car companies, offering rent-free facilities, and subsidizing each car sold. However, after NIO's bankruptcy, nearly 800 million of the 2 billion yuan invested by Yichun was lost.
- Meager industry profits: Price wars have reduced the overall profit margin in the automotive industry to just 3.4%, nearly half of the average for other industries. Many car models are sold below cost, and companies rely on financing to survive.
In short, while NEV technology is good (with a penetration rate of over 50%), capital moved too quickly, resulting in an oversupply of cars with no buyers, and the costs were borne by car owners, local governments, and small shareholders.
II. The AI Bubble: The "Super Balloon" That's Growing
The current scale of the AI bubble is much larger than that of the internet and NEV bubbles:
- Excessive capital investment: By 2026, the four major global tech giants (Google, Amazon, Meta, Microsoft) plan to invest $650 billion in AI infrastructure, which is 6.5 times the total VC investment during the peak of the internet bubble in 2000. Microsoft alone plans to invest $200 billion in building AI clusters, and Google will spend $180 billion on expanding data centers.
- Global adoption: China's 14th Five-Year Plan includes a 4-trillion yuan investment in computing power infrastructure; South Korea is investing 6.4 trillion yuan, and Japan plans to invest 4.3 trillion yuan by 2040. The EU is also funding chip development initiatives.
- Booming supply chain: AI server and GPU cabinet shipments have doubled, and Alibaba Cloud has reduced data center construction costs by 10% while expanding capacity.
All this money is being invested in "computing power infrastructure" (such as data centers and chips), but the question is: Is there really such a demand for computing power now? Just like with NEV companies that built factories without enough buyers, the result was an oversupply of cars.
III. The Financialization of Computing Power: The Dangerous Aspect of the AI Bubble
The AI bubble is more dangerous because of the "financialization of computing power"—turning potential future demand for computing power into assets that can be borrowed now:
- How it works: For example, NVIDIA partnered with financial institutions to create a $500 billion financing platform using future computing power contracts as collateral. A company called Volta obtained a $10 billion order based on a six-year computing power contract before even building its data center. CoreWeave could secure funding with just GPU purchase contracts.
- Risks: The internet bubble burned investors' equity, while the AI bubble is fueled by debt (bonds and private lending). According to the International Settlement Bank, U.S. tech companies will issue over $100 billion in AI-related bonds by 2025, and private lending has increased sevenfold. However, the bond market is becoming more cautious, with subscription multiples dropping from five to two.
- A weaker chain: The financialization chain relies on the assumption that future demand for computing power will continue to grow. If this does not materialize, the entire chain could collapse: idle data centers, bond defaults, and a burst in private lending, leading to global financial turmoil.
IV. The Cost of Bubble Bursting
The bursting of the NEV bubble affected car owners, local governments, and suppliers; the bursting of the AI bubble could have a much broader impact:
- Longer contagion chain: AI infrastructure involves chip manufacturing, data centers, and power supply, so any shortfall in profits could affect semiconductor, energy, and credit markets across industries.
- Passive arms race: Fearing lagging behind, companies are forced to expand production, leading to overcapacity and asset devaluation (e.g., idle data centers).
- Massive capital withdrawal: Trillions of dollars in passive funds are heavily invested in AI leaders. A market reversal could trigger a sharp decline.
- High market concentration: The U.S. stock market has risen largely due to AI companies; if AI declines, there may be no other sectors to support it.
For example, the bankruptcy of NEV companies affects a few thousand employees and tens of thousands of car owners, while the collapse of an AI data center could involve billions in bonds and private lending, impacting global investors.
V. The "Armor" of AI and Its Concerns
AI technology is real, but investments need to be cautious:
- Positive aspects: Leading companies like Microsoft and Google have stable cash flows and do not rely on external support. Generative AI has been implemented quickly (e.g., ChatGPT), and its penetration rate is faster than that of the internet. NASDAQ valuations (30 times) are lower than during the 2000 internet bubble (70-180 times).
- Concerns: Excessive investment in AI: Public funds hold nearly 40% of A-share market assets in the tech sector, more concentrated than in the consumer sector in 2020. The global AI market is highly interconnected (e.g., stock prices in China, the U.S., and South Korea move together). If the U.S. AI market declines, it could affect other countries.
The article concludes that good technology does not guarantee profitable investments. Just as NEV technology transformed the automotive industry, but most investors and car owners suffered, AI technology could also have significant consequences if the bubble bursts. Good technology is essential, but investments must be based on real demand, not just promising future prospects.
In summary: The NEV bubble was a small-scale experiment, while the AI bubble is a much larger one. Technology is genuine, but capital moves too quickly, and it's always ordinary people who bear the costs when bubbles burst. Advice: Avoid blindly chasing AI trends; invest based on real demand, not future narratives. Warning: The day the AI bubble bursts may be closer than you think. Reflection: Why do bubbles emerge with every technological revolution? Because capital always tries to "harvest the future" before it happens, but the future never follows a set script.