Institutional Capital Signals Sustained Three-Year Horizon for AI Infrastructure

The capital expenditure cycle underpinning the global expansion of generative artificial intelligence is projected to persist for at least three years. This assessment, provided by Jang Hyun-jun, head of equity management at Samsung Asset Management, frames the current market correction as a necessary stabilization phase within a broader, long-term secular growth trend.

Demand for data centers, high-performance semiconductors, and specialized power infrastructure continues to outpace initial forecasts. Jang Hyun-jun observed that the convergence of generative AI deployment and national-level strategic competition for computational dominance drives this sustained investment requirement. This structural demand persists even as the industry shifts from experimental applications to operational integration at scale.

Samsung Asset Management launched the Samsung U.S. AI Infrastructure Fund in June 2025 to capture value across these specific technical domains. The fund focuses on three primary pillars: advanced computing hardware, server architecture, and the physical data center infrastructure required to support massive model training and inference workloads. Since its inception, the fund has recorded a return of approximately 110%, with net assets reaching 25.5 billion won.

The fund employs a hybrid management strategy that deviates from standard ETF-only portfolios. By combining core AI-focused exchange-traded funds—such as those tracking U.S. semiconductor and optical communication networks—with targeted individual stock selection, the management team seeks to capture alpha that passive vehicles might miss. This approach allows for granular adjustments to portfolio weightings based on rapid shifts in hardware specifications and supply chain availability.

Technical agility remains a primary concern for the fund managers, who monitor the market on a daily basis. The strategy involves analyzing the underlying constituents of various ETFs to identify promising firms that may be underrepresented in broader indices. This process allows the team to augment their holdings with companies that are critical to the AI supply chain but are not yet fully captured by standard sector-specific ETFs.

The management team specifically targets firms involved in the cooling and power distribution segments of data center operations. As thermal design power requirements for next-generation GPUs continue to climb, the demand for specialized power delivery units and liquid cooling systems has become a critical bottleneck. Identifying these niche suppliers provides the fund with exposure to essential components that are often overlooked in broader, index-based investment products.

Risk management is integrated into the portfolio design to mitigate the inherent volatility of high-growth technology sectors. Semiconductors and power infrastructure components often exhibit significant price fluctuations, necessitating a more active analytical approach than traditional passive investment. Jang Hyun-jun emphasized that this rigorous evaluation process is essential for maintaining performance in an industry where the technical requirements shift as rapidly as the market conditions.

The initial industry consensus suggested that infrastructure spending would taper off as AI applications reached maturity. Contrary to these expectations, the demand for physical infrastructure has surged alongside the expansion of software capabilities. This sustained requirement for hardware, cooling, and power capacity indicates that the foundational layer of the AI stack remains a bottleneck for enterprise-scale adoption.

Market corrections in AI-related equities are interpreted as healthy breathers rather than signs of a structural downturn. These cycles of demand expectation, earnings verification, and profitability analysis are standard components of a maturing technology market. Future earnings reports are expected to resolve current concerns regarding financing and capital allocation efficiency.

The domestic semiconductor market in South Korea maintains a strong position due to the supplier-dominated nature of the memory sector. Because few global competitors can currently replicate the output and quality of Korean memory manufacturers, the industry remains a critical node in the global AI hardware supply chain. Analysts anticipate that the market will resume its upward trajectory once the current period of volatility subsides, with semiconductor performance serving as the primary catalyst for broader recovery.