The artificial intelligence wave is reshaping the valuation logic of the semiconductor industry, as AI-driven demand explodes, the memory chip market is undergoing a significant divergence. According to analysis, a research report released by Morgan Stanley on January 20 indicates that DRAM (Dynamic Random Access Memory) prices will continue to rise year-on-year in the first half of 2026, but this is not a feast where everyone benefits equally. The market is rapidly splitting into two distinct camps: the "winners" benefiting from AI inference and high-performance computing demand, and the "losers" constrained by macroeconomic headwinds and cost inflation.
Morgan Stanley analysis suggests that the technological inflation brought by AI is intensifying cost pressures across the supply chain, leading to a widening "wealth gap" between memory chip manufacturers and traditional hardware makers. Analysts point out that AI is not only driving robust demand for HBM (High Bandwidth Memory) and enterprise-grade SSDs, but is even causing persistent shortages in traditional memory products like DDR4. However, this rise in upstream memory costs is becoming a serious drag on profits for downstream PC and smartphone OEM manufacturers. In this "K-shaped" recovery, investors should pay close attention to companies positioned at the core of the AI supply chain. Morgan Stanley explicitly lists SK Hynix, Samsung, and Western Digital as "most favored" picks, citing their direct benefit from the AI-driven commodity cycle and the NAND super-cycle. Conversely, the market outlook remains pressured for traditional PC peripheral and OEM manufacturers unable to pass on rising memory costs to consumers, such as Acer and Logitech. The core logic behind this trend lies in the "crowding-out effect" of AI. As foundries and supply chains prioritize production for AI-related chips, capacity for non-AI sectors is being squeezed, leading to widespread supply tightness from high-end HBM to traditional DDR4. For investors, understanding this supply-demand mismatch is key to positioning in the 2026 semiconductor market.
Winners: Core Assets in the AI Wave Within the memory chip sector, AI is undoubtedly the most powerful catalyst currently. Morgan Stanley's report notes that suppliers with advanced memory technology and production capacity are experiencing a valuation re-rating, benefiting from the surge in demand for AI servers and inference. SK Hynix and Samsung are listed as top picks, with estimated upside potential of 13% and 14% respectively. This assessment is based on their dominant positions in the HBM market and improvements in the overall commodity cycle. Simultaneously, the NAND market is also entering a "super-cycle" supported by AI inference demand. Western Digital and Kioxia are viewed favorably due to their potential in the enterprise SSD segment. Analysis shows that AI storage demand is causing NAND shortages, and even NOR Flash supply constraints may persist into 2026.
Beyond direct memory chip manufacturers, semiconductor equipment (SPE) vendors are also beneficiaries of this boom. ASML Holding NV benefits from the increase in EUV layers, with a target price raised to 1,400 euros, implying 25% upside. Japanese equipment makers like Advantest and DISCO are also seen as high-growth stocks because their products are indispensable in the HBM manufacturing process.
Furthermore, among Greater China tech firms, Taiwan Semiconductor Manufacturing (TSMC) remains the undisputed core player. As the dominant foundry for AI logic chips, TSMC is not only planning aggressive expansion of its advanced packaging (CoWoS) capacity, but its 2026 capital expenditure and structural margin improvements are also robust. Morgan Stanley expects that with the volume production of Nvidia's Rubin chip in 2026 and the adoption of the N2 process for the Apple A20 processor, TSMC will maintain its dominance in advanced node technology.
Losers: Those Hampered by Cost Inflation In stark contrast to the exuberance of upstream memory suppliers, downstream hardware OEM manufacturers are facing severe challenges. Morgan Stanley analysts warn that memory cost inflation is seen as an industry-wide challenge, especially for consumer electronics brands lacking pricing power. Acer is listed as one of the "least favored" stocks, with a target price of just 20 New Taiwan Dollars, implying a potential 25% downside. The core reason is that rising prices for memory components cannot be fully passed on to end consumers, which will directly erode OEMs' profit margins. Similarly, HP and Dell face comparable headwinds; although they are attempting to enter the AI PC space, the near-term rise in memory costs remains a primary negative factor. This pressure also extends to the peripheral and radio frequency (RF) sectors. Logitech SA is expected to be impacted by memory inflation; these costs are not being offset by macroeconomic headwinds but are instead exacerbating operational pressures. RF giants Qorvo and Skyworks Solutions are also viewed less favorably due to concerns that high costs may suppress downstream hardware demand, which analysts believe will lead to further demand weakness.
Supply-Demand Mismatch: Not Just a Shortage of High-End Chips It is noteworthy that the impact of AI is not limited to high-end chips; it is creating a ripple effect across the entire memory supply chain. Morgan Stanley's research highlights a key phenomenon: the DDR4 shortage is expected to persist until the second half of 2026. This shortage is not due to a surge in demand for DDR4 itself, but rather because supply chain capacity is being crowded out by AI-related high-margin products like HBM and DDR5. As foundries prioritize production of high-profit AI chips, capacity allocation for traditional DDR4 and DDR3 is being reduced, leading to tight supply and consequent price increases. This benefits niche memory manufacturers like Taiwan's Winbond Electronics and Nanya Technology, as they possess stronger pricing power. Additionally, AI storage demand is also driving up spot prices for NAND wafers and module prices. With cloud service providers' (CSPs) capital expenditure projected to reach $632 billion in 2026 for data center construction dedicated to AI training and inference, vast storage resources will be consumed. This not only pushes up enterprise storage prices but also supports a recovery in consumer-grade NAND pricing, benefiting controller manufacturers like Taiwan's Phison Electronics Corp and Silicon Motion.
Long-Term Logic: The Rise of a Trillion-Dollar Market Looking ahead, Morgan Stanley reaffirms its long-term optimism for the AI semiconductor market. Nvidia's CEO has predicted global cloud capital expenditure will reach $1 trillion by 2028, and Morgan Stanley's data supports this trend, suggesting the global semiconductor market could hit the $1 trillion mark by 2030. Within this grand narrative, technological inflation, AI's self-cannibalization, and technology diffusion will be the three dominant market drivers. Although emerging AI models like DeepSeek demonstrate lower-cost inference capabilities, this does not alter the industry's fundamental reliance on high-performance computing and massive storage. For investors, the current strategy is clear: embrace upstream suppliers with core technologies whose capacity is being filled by AI demand, and avoid downstream hardware assemblers hampered by rising costs and a lack of ability to pass them on. Memory chip prices are not just a barometer of the industry cycle; they are the watershed distinguishing the market's winners from its losers.
Comments