[Real Investment Tech] Is It Safe to Invest in "Bubble" AI Stocks?

Altman Warns of "AI Bubble," Cautions Against Overvaluation of Startups
MIT Highlights "Lack of Generative AI Results," Stresses Agentic AI Efficiency
Alibaba in China Developing Semiconductors to Replace U.S. Nvidia
U.S.-China 'AI Decoupling' Accelerates, U.S. Advantage Likely to Continue

The surge in U.S. artificial intelligence (AI) stocks, led by Nvidia, which had been soaring without limits, has recently lost momentum. A series of negative factors have emerged, including Sam Altman's remarks last month about an "AI bubble," an MIT report on "insufficient AI performance," and, most recently, China's declaration of "independence in AI semiconductor chips."

What does the future hold for AI stocks, which have become a source of concern for Korean retail investors trading in U.S. markets? Stock market experts advise that concerns about AI are exaggerated, and that the "AI decoupling" between the U.S. and China was anticipated. They recommend viewing AI stocks from a medium- to long-term perspective.

The Recurring AI Bubble Debate

Let's turn the clock back to June of last year. Goldman Sachs raised the bubble argument, stating, "Despite massive investments, AI is not reaching where it needs to be." Sequoia Capital, a venture capital firm that has invested heavily in deep tech, also commented, "The gap between big tech companies' investment costs and their expected revenues is widening." At that time, AI-related stock prices temporarily stalled.

About a year later, last month, a similar event occurred. Immediately after Sam Altman, CEO of OpenAI and creator of ChatGPT, said in an interview that "this is similar to the dot-com bubble 20 years ago," the stock prices of the Magnificent 7 (Apple, Nvidia, Microsoft, Amazon, Alphabet, Meta, and Tesla) plummeted in the U.S. stock market.

Subsequently, it became known that MIT's NANDA (Networked Agents And Decentralized Architecture) reported that "companies have invested $3 billion to $4 billion in generative AI, but 95% have achieved no results." The market became engulfed in "AI bubble" sentiment.

However, it turned out that both the Altman interview and the MIT report were misunderstood. Jung Hoyoon, an analyst at Korea Investment & Securities, explained, "Unlike large language models (LLMs), which require massive capital and infrastructure, application development has a low barrier to entry. In his interview, Sam Altman warned that there is a bubble in the AI industry, as startups with only three or four developers and just an idea are attracting hundreds of millions of dollars in investment."

The MIT report's main point is not that AI utilization performance is low, but rather that the limitations of generative AI can be overcome by using agentic AI, which is seeing a growing number of users. Agentic AI does not require users to provide the full context every time it is used, and it learns through interaction with users, thereby increasing work efficiency.

LLM 'DeepSeek' Shock Followed by China's Declaration of AI Chip Independence

Earlier this month, The Wall Street Journal reported that "China's Alibaba is testing AI chips to replace Nvidia's H20." This came as another shock to the market, following the impressive efficiency demonstrated by China's "DeepSeek" in the LLM field earlier this year and now China's declaration to challenge the U.S. dominance in the chip sector.

At the end of July, the Chinese government pointed out security risks in Nvidia's H20 and urged Chinese companies to be cautious in purchasing it. At the same time, it is encouraging public data centers in Beijing, Shanghai, and other cities to source more than half of their semiconductors from Chinese companies.

Semiconductors produced by Chinese companies such as Huawei and Cambricon do not match Nvidia in performance and consume more power. However, China is developing LLMs that reduce computational burdens and is meeting power demands through large-scale investments in renewable energy. China is building an independent AI ecosystem, distinct from that of the U.S. Accordingly, investment bank Bernstein forecasts that the share of domestic AI chips in China's market will rise from 17% in 2023 to 46% this year, and to 55% by 2027.

U.S. AI Ecosystem-Related Stocks Worth Attention

Although China is challenging the U.S. in both AI software and hardware, many believe that the U.S. will ultimately prevail. The U.S. continues to invest more resources in artificial general intelligence (AGI). AGI is expected to provide overwhelming national competitiveness in research and development (R&D), economic growth, and security. In contrast, China, which lacks sufficient AI infrastructure, appears to be focusing on agentic AI that can deliver immediate results.

In the short term, U.S. AI-related stocks have faltered due to the "DeepSeek moment" highlighting China's efforts to catch up. However, the dominant analysis is that, in the long run, these stocks will trend upward. Kim Ilhyuk, an analyst at KB Securities, stated, "As long as concerns about the U.S. failing to achieve AGI do not intensify, and unless it is deemed better to pursue agentic AI as an optimization strategy, the U.S. will maintain its competitive edge for some time, justifying the high multiples of AI-related stocks."

Korean AI Policy Beneficiaries Also in the Spotlight

The rise of the Korean AI industry, which had been stagnant for the past three years, is also gradually becoming visible. In particular, the Lee Jaemyung administration is promoting the goal of making Korea one of the "top three AI powerhouses," drawing attention to related stocks listed on the domestic stock market.

With the official launch of the National Artificial Intelligence Strategy Committee, previously fragmented initiatives such as GPU supply, independent foundation model development, and industrial AI expansion projects will now be managed under a unified system, strengthening policy execution. In the future, a concrete AI action plan will be released, followed by implementation tasks such as industry-specific AI competitions in manufacturing and healthcare, and the third round of national AI computing center projects. The structure will operate in full swing, moving from governance (National AI Strategy Committee) to master plan (AI action plan) to implementation tasks (industry-specific AI and computing center projects).

From an investment strategy perspective, investors can take a dual approach by targeting both direct beneficiaries and sectors that will benefit indirectly. Park Kihun, an analyst at Korea Investment & Securities, analyzed, "Direct beneficiaries include Naver, Kakao, and NHN, which receive cloud infrastructure, as well as SK Telecom, NCSoft, and LG Group stocks, which are part of the national AI foundation model development team. Once AI computing center and industry-specific AI competitions begin in earnest, it will also be worth paying attention to Samsung SDS, Saltlux, Lunit, Konan Technology, and Synapsoft, which are expected to participate."