Summary of Key Points
This article discusses the current global AI investment boom and highlights the hidden risks of a bubble: The AI hardware (infrastructure) sector is being excessively hyped due to a "selling shovels" logic, where providers of necessary components see significant profits. However, the application side has not yet established a stable profit model, raising doubts about the sustainability of capital expenditures. By drawing on historical examples (such as Nortel and Cisco during the internet bubble), the article warns investors that technological revolutions do not guarantee returns on investment, and blindly chasing hardware bubbles can lead to substantial losses. It ultimately advises investors to adopt a strategy of "selecting, waiting, and holding," focusing on platform companies with long-term competitive advantages rather than short-term bubble targets.
Who Is Really Making Money in the AI Gold Rush?
If we compare AI to a gold rush, the real beneficiaries are not the "gold miners" (application companies) but the "shovel sellers" (hardware infrastructure manufacturers):
- Shovel sellers are making huge profits: Companies like NVIDIA (GPU chips), TSMC (chip manufacturing), and Zhongji Xuchuang (optical modules) have reaped substantial gains by supplying computing power to AI firms. Data shows that in 2026, NVIDIA accounted for nearly one-third of the global AI industry's profits, with U.S. and Korean companies together accounting for 84%.
- Gold miners haven't made money yet: Giants in the application space, such as Amazon and Microsoft, have invested billions in hardware but have not seen stable revenue from consumer subscriptions or business services, and some are still losing money. This indicates that while AI applications are still in the conceptual phase, the hardware has entered a "realization" stage—however, this realization is based on massive spending by downstream companies. If these downstream firms fail to generate profits, demand for hardware could plummet.
The Risks of the Current Hardware Bubble: History Speaks Loudly
The current market frenzy for AI hardware is similar to the 2000 internet bubble, with several examples highlighting the dangers:
- Nortel: A leading global fiber optic company at the time (equivalent to NVIDIA today) with a $250 billion market value. When downstream internet companies stopped investing, Nortel's fiber sales plummeted, leading to bankruptcy and massive investor losses.
- Cisco: Its valuation exceeded its future earnings potential; despite billions in profits, its $400 billion market value was the sum of expected earnings for decades. After the bubble burst, its market value dropped by 75%, and it never regained its peak for 20 years.
- Microsoft: A platform company that survived the crisis by transitioning to cloud services (Azure) and AI. This shows that while hardware technology evolves rapidly and can be obsolete, companies capable of building ecosystems are more resilient.
The problem with current AI hardware is that supply is dominated by a few oligarchs like NVIDIA, and there is no stable demand from downstream applications. If these giants cannot turn profits, hardware demand will collapse.
The Three Layers of the AI Industry: Where Are the Greatest Opportunities?
The AI industry has three layers, each with different risks and opportunities:
1. Infrastructure layer (chips, computing power, optical modules): This is the hottest sector but also the most risky. Technology evolves quickly (next year might bring more advanced GPUs), and companies that fail to keep up will be eliminated. Valuations are already sky-high (e.g., 240 times on the STAR Market), overestimating future potential.
2. Platform layer (cloud computing, large model frameworks): Often underestimated but crucial. Companies like Microsoft's Azure and OpenAI's GPT frameworks can transform technology into ecosystems and lock in users. They have the capital and data barriers to remain competitive.
3. Application layer (robots, AI-enabled devices): Great potential but with low survival rates. Some large model companies may have revenues of only $700 million but market values in the tens of billions, eventually being eliminated due to unclear business models (similar to many .com companies during the internet bubble).
How to Survive a Bubble Burst?
The article recommends a traditional investment strategy: "select, wait, and hold":
- Select: Choose companies with competitive advantages, such as data barriers (e.g., WeChat's user base) or ecosystem strengths (e.g., Apple's App Store), rather than those that rely on spending money to build computing power.
- Wait: Don't rush to buy; wait until the bubble bursts. For example, after the internet bubble, Microsoft and Google's stock prices dropped to reasonable levels, providing buying opportunities. Hardware valuations are currently too high; it's better to hold cash.
- Hold: Invest through broad-based indices for long-term exposure. The NASDAQ index, which includes many global tech leaders, helps diversify risks.
In short: **Avoid buying the most overheated hardware stocks now; wait for the bubble to burst and then invest in platform companies or indices that will survive.
How to Invest in AI Today: Return to Common Sense
The article emphasizes that while AI will transform the world, investing is not about betting on big gains:
- Don't believe the myth that missing an opportunity is worse than the risk of a bubble: Those who bought internet stocks at the end of 1999 waited 15 years to recover their investment. The same may apply to AI hardware investments.
- Focus on cash flow: No matter how appealing the story, companies must generate profits (free cash flow) to sustain themselves; those that only spend money without making a profit will be eliminated by the market.
- Look for future blue chips: Invest in companies with platform ecosystems, data barriers, and continuous R&D capabilities—these are the ones that can withstand market fluctuations.
In conclusion, AI is a long-term trend, but the current hardware bubble is a short-term phenomenon. The real opportunities lie in platforms and companies that create lasting value, not in short-term speculation. By drawing on historical lessons, investors can avoid being misled by hype and return to sound investment principles.