Interest Rates Can Suppress AI Stock Prices, But Not Demand: The Truth About Semiconductor Shortages | Kim Jang-yeol, Unistory Asset Management Research Center Head

Summary

Kim Jang-yeol argues that current market fears over AI monetization are mostly an OpenAI-specific problem, while token demand, KV cache memory growth, and Big Tech customer-funded capex keep AI infrastructure demand intact. He sees recent Samsung Electronics and SK Hynix weakness as price noise rather than fundamental damage. The main near-term risk is rates: 10-year Treasury yields near 5% or 30-year yields near 5.7-5.8% could trigger sharp U.S. and Korean equity declines.

• The speaker identifies liquidity and semiconductor demand concerns as the two main market variables.
• He argues AI monetization fears are concentrated in OpenAI/Oracle, while Google, Anthropic, Amazon, Nvidia, Nebius, and CoreWeave are on stronger footing.
• Agent AI token growth and KV cache memory expansion support continued HBM and DRAM demand.
• Neocloud firms have customer prepayments and large backlogs, reducing AI capex financing risk.
• Samsung SDS is highlighted for AI data center value and long-term robot orchestration potential.
• Korean semiconductor equipment/materials are expected to benefit from future domestic and U.S. capex.
• He recommends balancing Korean AI exposure with non-AI trading sectors such as construction, shipbuilding, and cosmetics.
• The key rate thresholds to watch are 5.0% on the 10-year Treasury and 5.7-5.8% on the 30-year Treasury.