Summary of Key Points
Recent adjustments in the AI market have sparked debates about whether a "bubble" has burst. However, the article argues that AI represents a transformation fundamentally different from previous technological revolutions, such as the Internet or the Industrial Revolution. It features significant involvement from national governments, the creation of a new dimension of "general intelligence," and a much faster pace of demand growth. Currently, there is a shortage of AI hardware, with no signs of overcapacity in infrastructure. In the future, demand will primarily come from the business sector (B2B), with multiple areas of growth rather than relying on a single "killer app." The supply side is likely to see a pattern of hierarchical oligopoly. The article also emphasizes that technological revolutions rely not only on supply but also on domestic demand and social mechanisms to avoid repeating the mistakes Japan made by missing out on the Internet application boom.
Why AI Isn't a "Replica of the Internet Bubble"? It's Different from Previous Revolutions
Many people compare AI to the Internet bubble of 2000, but the article points out that the two situations are entirely different:
1. Greater Government Involvement: AI is a focal point of competition among major nations, with countries not only imposing technology embargoes but also directly investing in research, government procurement (e.g., purchasing AI systems), and building data centers. If a bubble does form, it will be more "solid" due to this support.
2. **The Addition of a New Dimension of "General Intelligence": Previous revolutions (such as electricity and the Internet) addressed specific issues (energy, communication), while AI breaks the monopoly on "general intelligence"—machines can now perform tasks like translation, coding, and data analysis, opening up a wider range of applications.
3. Faster Demand Growth: With the Internet, infrastructure (fiber optics, servers) was built first, followed by applications; with AI, demand and infrastructure develop simultaneously, and sometimes even faster. For businesses, adopting AI to improve processes (e.g., using AI for coding) is cost-effective and yields quick results, so there is no issue of "lagging demand."
Is There an "Overcapacity" in the Current AI Market? The Answer is: Not Yet
There are concerns that AI might experience overcapacity in infrastructure, similar to the Internet bubble, but the article provides evidence that this is not the case:
1. Hardware Shortage: There is still a severe shortage of AI hardware (GPUs, storage), and prices are rising sharply. The decline in hardware stock prices in July was due to concerns about the slowdown in price increases, not an oversupply.
2. Capacity Release Lags: Hardware production capacity is not expected to increase until 2027-2028, but by then, demand will likely be much higher (for example, with more businesses adopting AI).
3. Continuing Demand Growth: The amount of data processed by AI (Token demand) is growing faster than supply. Open-source models are becoming more affordable, making them more accessible to businesses, thus driving demand even further.
Where Will the Next Wave of AI Demand Come From? Not from Single Apps, but from Multiple Business Areas
The Internet's demand was driven by "killer apps" like WeChat and TikTok, but AI is different:
1. Business Sector as the Driver: For instance, Palantir in the U.S. has already made money from its AI services, and Chinese companies are also advancing AI deployments to improve business processes, which will be a key driver of growth in 2027.
2. Multiple Areas of Growth: AI is being applied in fields such as biomedicine (drug discovery), gaming (AIGC), and research. Each industry will develop its own AI applications, without the need for a single dominant platform.
3. Self-Accelerating Demand: The more businesses use AI, the more efficient they become, which in turn drives more adoption, creating a "flywheel effect." As long as current demand is genuine, it will continue to grow.
What Will the Future of AI Look Like?
The article predicts the following landscape for AI:
1. Supply Side: A hierarchical oligopoly, with a few companies dominating core areas such as chips (GPUs), large models, and computing power centers (e.g., NVIDIA, OpenAI), due to economies of scale. Standardized hardware (e.g., regular servers) will face more competition.
2. Demand Side: The business sector (B2B) will lead demand, as companies are willing to invest in AI for tasks like risk management and process optimization. Consumers (C2C) are less likely to pay for AI services.
3. Limited Opportunities for Consumers: AI may not significantly change consumers' lifestyles like the Internet did, unless it creates new leisure time or innovative consumer-facing products (e.g., more intelligent AI assistants). However, its impact on consumers will mainly be indirect through the business sector.
Macro Insights: Technological Revolutions Require More Than Just Technology
The article concludes that being technologically advanced is not enough; domestic demand and social mechanisms are also crucial:
- Japan's Lesson: Japan had strong semiconductor technology in the 1970s-1980s but saw its domestic market shrink and lacked the investment necessary to catch up with the Internet revolution, resulting in missed opportunities.
- The Key to the AI Era: It's essential to ensure that consumers have the means to spend and that businesses are vibrant. A mechanism that transforms technological progress into social benefits is needed; otherwise, even leading technology may fail due to insufficient demand.
In summary, AI is not a bubble, and its future is promising. However, to fully realize its potential, it will require the cooperation of the entire society.