AI market differs from dot-com bubble due to hyper-scaler leadership

Economics Master Kwak Soo-jong argues that the current AI market is a structural competition rather than a bubble, noting that unlike the investor-driven dot-com era, today's market is led by financially sound hyper-scalers. While hedge funds reached $13 trillion by 2025, the capital required for the AI field is predicted to reach $10.3 trillion by 2031.

Speculation regarding an "AI bubble" is intensifying due to concerns that technological expectations may not translate into actual profits. However, an analysis suggests that comparing the current AI craze to the dot-com bubble that shook the market in the late 1990s is unreasonable, as the leading entities, the nature of capital, and the structure of infrastructure are fundamentally different.

According to the analysis provided by Economics Master Kwak Soo-jong, the most significant difference between the dot-com bubble and the AI market lies in "who leads the market." During the dot-com bubble, the market was driven by expectations and illusions created by investors, such as hedge funds, rather than by clear corporate leadership. At that time, the scale of hedge funds was approximately $3 trillion to $5 trillion, and indiscriminate investment flowed into startups and venture companies with unclear identities. "The dot-com bubble was a period of indiscriminate investment within illusions created by investors, and there were no companies that could be called leading entities," Kwak explained.

In contrast, the current AI market is led by "hyper-scalers"—big tech companies with massive capital power. Having experienced the failures of the dot-com era, these companies operate with secured financial soundness. "Because massive capital is required, the companies themselves must possess the money, and it is only possible for hyper-scalers such as the M7 to act while maintaining financial soundness," Kwak analyzed. "The nature of the funds is entirely different from the past, which was investor-centered." However, he added, "The problem is that even the M7 currently lack money."

IP Address Wars vs. Data Center and Power Infrastructure Wars

A stark contrast also appears in the nature of fundamentals and infrastructure. The core infrastructure during the dot-com bubble was a simple digital information base at the level of fiber-optic cables. The market was so primitive and speculative that securing website addresses (IP addresses) alone could generate millions of dollars in profit. "It was a time when one could earn millions of dollars depending on who registered an HTTP address first," Kwak recalled, noting how primitive and undeveloped the market was. Vague expectations that "profits will come later" dominated without specific considerations for return on investment (ROI).

However, the fundamentals of the AI era are much more complex and massive. A giant infrastructure system centered on data centers—intertwined with power, semiconductor chips (GPU, NPU, TPU, CPU, etc.), memory, and data networks—is essential. "In the AI era, everything from data centers and power to semiconductors and memory is trending toward becoming infrastructure," Kwak analyzed. "Only by possessing such infrastructure is the qualification granted to major companies like the M7."

The monetization models are also different. While external metrics such as subscriber counts were important in the past, the current battle is about who can maintain truthful technological capabilities and generate actual profits until the end. Having passed through the 2008 financial crisis and the 2000 dot-com bubble, the market now possesses a sophisticated verification system that examines ROI and price-to-earnings (P/E) ratios. "We have reached a period where it will be determined who told the truth consistently until the end and who was ultimately right," Kwak said.

China's Low-Cost Offensive and the Task for South Korean Companies

Kwak diagnosed that the AI market has entered a "structural competition stage" rather than being a "bubble." In particular, the biggest variable facing U.S. big tech companies is the low-cost, open-source offensive from China. If China releases high-performance AI at a low cost through models like DeepSeek, U.S. Large Language Model (LLM) programs may struggle to attract consumers. "It seems China will extract data created in the United States and manage only the most frequently used essence," Kwak noted. "The core of the competition will be whether China can be defeated through open source and how the advantages of infrastructure construction can be utilized."

The scale of this capital war is beyond imagination. With the scale of hedge funds reaching $13 trillion as of 2025, the capital required for the AI field is predicted to reach as much as $10.3 trillion by 2031. "The structure of AI is such that capital must continue to be poured in like 'pouring water into a bottomless pit,'" Kwak forecasted. "Due to this massive scale of capital, it does not seem easy for South Korean companies to catch up."

Ultimately, the AI market is expected to move in a direction where the nation or company that preempts technological standards sets the rules. "Technology must have standards, and there must be rules required when using those standardized technologies," Kwak said. "The responsible parties for rule-setting and standardization will be the companies or nations that possess that technology and its originality."

Bubble Collapse Scenarios: Great Depression Risks and Ordered Reorganization

If the AI market were truly a bubble centered on indiscriminate investment like the dot-com bubble, its collapse could immediately transfer to the real economy like the 2008 financial crisis, leading to a Great Depression. "Bubbles are bound to burst, and when they do, a Great Depression-like situation is inevitable," Kwak warned. "This is because the collapse of capital markets immediately ignites the real economy within days or even hours." He mentioned that if a Great Depression-like situation occurs, the KRW-USD exchange rate could soar past 2,000 won to reach 3,000 won.

However, Kwak maintains the position that it is difficult to view AI as a bubble in the same form as the dot-com bubble. Because current AI operates based on clear capital scales and infrastructure, he believes that even if the market undergoes an adjustment, it is more likely to be reorganized in an orderly manner through established mechanisms such as mergers and acquisitions (M&A), rather than through the crude and irregular collapse seen in the past. "The process of AI being organized will not be crude and irregular like the dot-com bubble," Kwak explained. "Wouldn't it be organized or undergo M&A following a certain order, mechanism, and frame?"

Source: original video (YouTube)