Samsung Electronics and SK Hynix just posted earnings that put them in the same league as NVIDIA, Microsoft, and Apple, yet global investors are still undervaluing them by roughly 50 percent. In the first quarter of 2026, Samsung posted 57.2 trillion won in operating profit, while SK Hynix hit 37.6 trillion won. Combined, their earnings exceeded TSMC's (Taiwan Semiconductor Manufacturing Company), the world's most valuable chipmaker, yet their combined market value was only around 70 percent of TSMC's. The reason? Most investors still see Korean memory companies as cyclical commodity players, not as structural pillars of artificial intelligence (AI) infrastructure. That perception gap is the story.

What's Actually Changed in the AI Chip Market?

The shift from AI training to AI inference is reshaping which chips matter most. Training large language models (LLMs) requires massive computing power, but inference, the process of running those models in real time for millions of users, requires something different: memory bandwidth, capacity, and power efficiency. When a company like OpenAI or Google serves ChatGPT or Gemini to millions of simultaneous users, it needs to load tens of billions of parameters into GPU memory. That memory is called HBM, or high-bandwidth memory.

Here's where Korea becomes indispensable. Samsung and SK Hynix combined account for well over half the global HBM market. Looking at DRAM (dynamic random-access memory), the standard memory used in servers, the two companies are core suppliers covering the majority of global share. The Cloud Big 4, meaning Microsoft, Google, Amazon, and Meta, are expected to push their combined capital expenditure past 700 billion dollars in 2026. Not all of that goes to GPUs. Data centers, power, networking, cooling, and memory all grow together. Within that ecosystem, HBM and high-capacity server memory have established themselves as non-negotiable components.

Why the Valuation Discount Doesn't Make Sense Anymore

For decades, investors treated memory companies as cyclical. Prices would surge when demand exceeded supply, then crash when supply caught up. The market learned this pattern and applied cyclical multiples, meaning investors would only pay a fraction of earnings for these stocks because they expected those earnings to disappear in the next downturn. What is happening now is fundamentally different. Long-term supply agreements spanning three to five years are being signed between Korean memory makers and Big Tech companies. This is an evolution toward the production model that TSMC pioneered, where orders are locked in years in advance rather than fluctuating with spot market demand.

This structural shift is dampening earnings volatility and expanding visibility. Investors are starting to recognize this, and capital is flowing into Korean semiconductor stocks, but valuations have not fully caught up. A gap remains between the cyclical multiples applied in the past and the multiples appropriate for companies with stable, long-term demand. The narrowing of that gap is the investment opportunity that global investors are now beginning to see.

How to Understand Korea's Hidden AI Semiconductor Supply Chain

• Memory and Infrastructure Layer: While the U.S. market focuses on the chip and cloud layer of AI, Korea sits on the memory and infrastructure layer. Both sit on the same AI capital expenditure, but they give access to different tiers of the supply chain, allowing investors to pick individual bottlenecks rather than betting on AI as a whole.
• Equipment and Component Suppliers: Below Samsung and SK Hynix, dozens of companies form the supply chain across equipment, materials, test, packaging, power, and optics. Each is connected to entirely different demand drivers. Niche companies connected to AI infrastructure bottlenecks exist as independent listed entities, from HBM test sockets to ultra-high-voltage transformers.
• Governance Discount Factor: Chaebol governance structures, disregard for minority shareholder rights, low dividend payout ratios, and opaque decision-making create a valuation discount that does not apply equally to every stock. Large conglomerates carry different types of discount than mid-cap and small-cap equipment companies, so investing in Korea's AI semiconductor space requires understanding why stocks are cheap and identifying which can escape that discount.

The Korean semiconductor ecosystem breaks down into four distinct categories, each moving at different timings and in different directions depending on which demand they connect to. Stocks that move when Samsung's capital expenditure moves include suppliers to memory, HBM4, foundry, and advanced packaging. Stocks that ride along when SK Hynix's HBM sells include TC bonding, test, and substrate suppliers. Test complexity increases non-linearly as stack count rises, and the market has not yet fully priced this in. Stocks that benefit when AI data centers get built include transformer suppliers, which represent the bottleneck before the server even turns on. Finally, AI server components and optical interconnects form their own category with distinct demand drivers.

Why Global Investors Are Suddenly Paying Attention to Korea?

On May 6, 2026, the KOSPI, South Korea's main stock index, broke through 7,000 for the first time. Samsung Electronics joined the trillion-dollar market cap club. SK Hynix hit an all-time high. Foreign investors net-bought over 3 trillion won in a single day. The numbers matter less than the direction. Global investors are turning their attention to Korea because the country's position in the AI semiconductor supply chain is being re-rated from a cyclical memory trade to a structural pillar of AI infrastructure.

The tightening of U.S. restrictions on China is also pushing up the global share of Korean companies, reinforcing this structural advantage. As Western tech companies face limits on selling advanced chips to China, they are diversifying their supply chains and deepening relationships with trusted partners like South Korea. This geopolitical shift, combined with the fundamental shift from training to inference in AI workloads, creates a rare moment where valuation and fundamentals are misaligned in Korea's favor.

Most investors still stop at Samsung Electronics and SK Hynix when analyzing Korean semiconductors. But the real opportunity lies in understanding the supply chain below these two giants. The companies that supply test equipment, substrates, power components, and optical interconnects are positioned to benefit from the same AI infrastructure buildout, often with less analyst coverage and lower valuations. For investors seeking pure exposure to specific AI infrastructure bottlenecks rather than broad bets on AI as a whole, Korea offers something rare: a tightly mapped ecosystem of independent listed companies, each connected to a different constraint in the AI supply chain.