Summary of Key Points
The South Korean government, leading the Samsung and SK Group corporations, plans to invest $1.3 trillion (approximately 2000 trillion Korean won) over the next decade to heavily support the domestic semiconductor and storage industry, with a focus on high-performance storage solutions required for AI, such as HBM (High Bandwidth Memory). This is not a spontaneous expansion by the companies but a strategic bet at the national level—betting that the demand for high-performance storage will persist for more than a decade, thereby consolidating South Korea's core position in global AI infrastructure. However, this bold move also faces the cyclical challenges of the storage industry: Can AI change the length of these cycles, or will it completely eliminate them? The answer remains unknown.
Detailed Analysis
1. Is $1.3 trillion really that much?
$1.3 trillion is no small amount:
- In comparison to national GDP: South Korea's GDP is expected to reach $1.87 trillion by 2025; this investment is nearly 70% of its annual economic output, equivalent to investing most of the country's annual production in semiconductors.
- In comparison to corporate assets: Samsung has a market value of about $1.34 trillion, and SK Hynix of about $1.2 trillion; together, they are committing nearly half of their total market value over a decade of expansion. In simple terms, this is a major gamble involving the two largest financial groups in South Korea, backed by the government.
2. Why has AI turned storage from a “supporting role” to a “central role”?
Previously, storage was a “low-profile supporting role” within the semiconductor industry, with strong cyclical patterns (increased demand → price increases → expansion → oversupply → price drops → and so on). But with the advent of AI, everything has changed:
- Storage has become a bottleneck for AI computing power: GPUs are the “brains” of AI, but HBM is the “blood vessels”—without fast enough memory, even the most powerful GPUs cannot perform effectively. The larger the models and the more frequent the computations, the greater the demand for high-bandwidth, high-capacity storage.
- Orders booked years in advance: Global AI chip manufacturers (such as NVIDIA) and cloud giants (like Amazon and Google) are competing to order HBM, with orders from Samsung and SK Hynix already scheduled for 2026 and beyond, driving their performance and valuation upwards.
Storage is no longer a cyclical supporting element; it has become a critical, essential part of the AI supply chain.
3. What are the South Korean government’s motivations behind this investment?
This investment was not decided by the companies alone but is a strategic initiative led by President Lee Jae-myeong:
- To consolidate global storage dominance: South Korea is already a leader in the storage industry (Samsung and SK Hynix control 70% of the global DRAM market and 40% of the NAND market). In the AI era, storage has become a core component of AI infrastructure. By taking the lead, the government can ensure the allocation of resources (such as land for factories, tax reductions, and subsidies), thereby strengthening its competitive advantage.
- To gain influence in the AI era: The AI race is essentially a competition for infrastructure, and storage is an indispensable part. With this investment, South Korea aims to signal to the world: “AI cannot do without our storage; we want to occupy a key position in the AI era.”
4. Can this gamble be successful? Will the cyclical nature of storage disappear?
The biggest question is whether AI can eliminate the cyclical patterns in the storage industry:
- Optimistic view: HBM production is complex and requires advanced packaging technology, so capacity expansion will be slow. Even with increased production by major manufacturers, supply-demand shortages may persist for 3–5 years. The rapid growth in AI demand, along with model iterations and data center expansions, will continue to drive demand for storage, with no immediate ceiling in sight.
- Risk view: If future AI demand slows down and new capacity is released quickly, it could lead to oversupply and price drops. While AI may extend the cycle (e.g., from 3–5 years to 5–10 years), it may not completely eliminate cyclical patterns. South Korea’s bet is that AI demand will last for a decade, and they are investing trillions in hopes of creating a “super cycle.” The outcome can only be determined by time.
5. What impact will this have on ordinary people?
Although it may seem distant, the implications are significant:
- Cheaper AI products: If storage capacity increases, the cost of AI servers and large-scale models could decrease, making AI tools more affordable for consumers.
- Stabilized South Korean economy: The storage industry is vital to South Korea’s economy; this investment can boost employment and technological advancement, enhancing the country’s competitiveness in the global market.
- Intensified tech competition: The United States and China are also investing in the storage industry, so this move will intensify the global competition for AI infrastructure, ultimately benefiting the entire tech sector.
In summary, this $1.3 trillion investment represents South Korea’s all-in commitment to the AI era, betting that storage can become a decisive factor in the AI race. Regardless of the outcome, it marks a fundamental shift in the storage industry from a cyclical supporting role to a central component of AI infrastructure.