Summary of Key Points
Over the past few weeks, the South Korean stock market has experienced extreme fluctuations characterized by circuit breakers, sharp rises, and subsequent plunges. Behind these events lies a deep-seated flaw in the country's AI industry structure: an over-reliance on HBM memory chips (with Samsung and SK Hynix dominating 80% of the global market), coupled with a lack of control over the core components of the entire AI ecosystem, such as large-scale models and computing power. The intensifying AI competition between China and the United States has fueled market expectations for HBM, leading to a bubble that burst and caused widespread losses among retail investors, social crises (including suicides). This phenomenon reflects a common dilemma for mid-power countries in the AI era: they must either bet on a single strength and create an overheated bubble or become dependent on the Chinese and American systems, thereby losing their technological sovereignty. China's open-source AI approach may offer these countries a third alternative that does not force them to take a side.
I. Why Is South Korea in a Critical Situation? It Only Has “Chip Parts,” Not a Complete “AI System”
South Korea's advantage in the AI field is akin to “walking on one leg”: it possesses some of the world's best HBM memory chips (essential for running large-scale AI models), but it lacks the “brain” (large-scale models) and the “heart” (computing power chips like NVIDIA GPUs) of an AI system.
- Relying on a Single Component: HBM is a critical link in the AI supply chain, but it is merely a component, not a finished product. South Korea has bet its entire national AI strategy on HBM, tying its stock market, policies, and retail investor wealth to this single sector, thereby exponentially increasing the risks.
- Two Companies Dominating the Market: Samsung and SK Hynix once accounted for 60% of the South Korean stock market's weight, and the index's movements were largely determined by their performance. If investors lose confidence in AI-related capital spending (for example, if companies stop buying HBM aggressively), their stock prices could plummet, dragging down the entire market.
- Lack of Control: The demand for HBM (e.g., by large American model companies) and its price (determined by global data center investment trends) are influenced by Chinese and American giants. South Korea is at the mercy of these companies; any change in their orders can have severe consequences.
II. Why Does South Korea Suffer from the Consequences of China-US AI Competition?
The volatility in South Korea's stock market is not an accident but a result of the spillover effects of this competition:
- Exaggerated Expectations: The competition between China and the US has led to inflated expectations, with both sides producing stronger (closed-source) and cheaper (open-source) AI models. This has created the notion that AI data centers will expand indefinitely, driving demand for HBM. South Korean companies have even predicted a “super cycle” for HBM in 2026, prompting retail investors to buy heavily on leverage, which further inflated the bubble.
- The Chain Reaction of a Bursting Bubble: Once doubts arise about AI investment (e.g., if NVIDIA's chip shipments fall short of expectations), HBM prices plummet, triggering circuit breakers and causing retail investors with leveraged accounts to lose all their investments or even owe money to brokers.
- Anxieties of Mid-Power Countries: Fearing becoming technological dependencies, these countries (like South Korea) are forced to bet heavily on HBM. This concentration of resources and capital in a single sector can lead to crises.
III. The Dilemma for Mid-Power Countries: Bet Big or Lose Autonomy
South Korea's situation is a microcosm of the challenges faced by all mid-power countries in the AI era:
- Option A: Betting on a Single Strength Leading to a Bubble: Countries may focus on a single strength, such as HBM (South Korea), semiconductor materials (Japan), or lithography machines (Netherlands). However, this can create bubbles that are vulnerable to collapse, causing severe economic and social impacts.
- Option B: Becoming Dependent on Chinese and American Systems: Some countries, like the UAE’s G42 and Saudi Arabia’s Humain, may seek investments from companies like Microsoft or make trade-offs (e.g., buying NVIDIA chips while avoiding Huawei equipment). While they gain access to technology, they lose control over their own industrial paths and suppliers, becoming technological puppets.
In either case, these countries risk losing their influence in the AI landscape, either by being limited to supplying components or merely acting as customers with no say in defining technical directions or industry standards.
IV. Could China’s Open-Source AI Provide a Third Path for Mid-Power Countries?
China’s open-source AI initiatives (such as Alibaba’s Qianwen and Kimi K3) could offer a viable alternative:
- Building Their Own AI Infrastructure: Open-source models provide a foundation that mid-power countries can use to develop their own industry-specific applications (e.g., in smart manufacturing or automotive AI). This diversifies their reliance on HBM and reduces risks.
- Preserving Technological Sovereignty: Unlike closed-source models from the US, open-source models can be customized and deployed locally, allowing these countries to retain control over their technology.
- Negotiation Power: With their own AI ecosystems, mid-power countries can negotiate better terms with companies like NVIDIA or large American model providers, rather than being at their mercy.
While this approach is not a panacea, it does at least give them the flexibility to avoid the stark choices between relying on bubbles or becoming dependent on others. South Korea’s experience serves as a reminder that without exploring alternative paths, the next country to face a crisis could be them.