Diversification Buffers Korean AI ETFs as Memory Stocks Stumble

Korean asset managers are witnessing a sharp divergence in the performance of AI-focused active ETFs tracking US markets. Over a one-month period, a performance gap of up to 13 percentage points has emerged, revealing a growing divide between funds diversified across the AI value chain and those heavily concentrated in semiconductor memory.

According to the Maeil Business Newspaper, the ACE U.S. AI Tech Core Industry Active and RISE U.S. AI Tech Active funds recorded losses of 10.09% and 11.94%, respectively. These funds outperformed their peers by diversifying into Big Tech and critical AI infrastructure, including holdings in Vertiv Holdings, Oklo, and Cameco. In contrast, the TIME Global AI AI Active (-14.72%) and TIGER Global AI Active (-22.70%) funds suffered steeper declines, driven by high exposure to memory and storage stocks such as Kioxia, AMD, SanDisk, and Western Digital.

The Memory Decoupling

This volatility arrives as the semiconductor market experiences a contradiction. TrendForce reports that while the pace of price increases for DRAM and NAND has slowed, prices continue to rise. This has created a 'decoupling' effect where stock prices for memory providers fall even as the real-world price of the hardware climbs. This suggests that current stock valuations may have been overextended, triggering a correction despite solid underlying earnings.

Meanwhile, the broader Big Tech sector has shown resilience. Market capitalization for Big Tech increased by $1.5 trillion, fueled largely by the growth of cloud services from Alphabet, Amazon, and Microsoft. This shift indicates that the market is rewarding the platforms providing the cloud scale necessary for AI deployment over the companies providing raw storage components.

A Shift in the AI Trade

This divergence highlights a fundamental shift in investor sentiment. The 'AI trade' is no longer a monolithic bet on GPUs or memory; it has evolved into a broader play on the entire value chain. Success now depends on balancing high-growth cloud services with the physical infrastructure—such as power and nuclear energy (SMRs)—and customized silicon (ASICs) required to support massive compute loads.

Park Seung-jin, head of Hana Securities' overseas stock analysis division, notes that active ETFs can be more advantageous than individual stock picking because fund managers can preemptively adjust weights based on emerging issues. This flexibility is becoming critical as the market moves away from pure-play hardware.

Outlook for Investors

Despite the recent dip, some analysts see a buying opportunity. The U.S. investment magazine Barance suggests that memory for AI servers remains a long-term opportunity for investors to buy on dips, as supply continues to be constrained. Investors will likely watch whether the 'decoupling' of memory prices and stock values stabilizes or if the trend toward infrastructure and cloud services continues to dominate the AI landscape.