Samsung and SK Hynix Capex Plans Put Applied Materials and Lam Research in Focus

Korean Foundry Capex Returns to Focus

Applied Materials is drawing renewed investor attention following capital expenditure announcements from Samsung Electronics and SK Hynix, the two largest DRAM and high-bandwidth memory manufacturers globally. Both Korean companies are expanding capacity to meet demand for HBM chips, which serve as the dedicated memory layer in AI accelerators and have become a structurally constrained component in the AI compute supply chain.

SK Hynix is the dominant supplier of HBM3E, the current-generation high-bandwidth memory used in Nvidia's H100 and H200 GPU systems. Samsung has been working to qualify its HBM product after losing share to SK Hynix and is investing to close that gap. The combined capital programmes from both companies represent a substantial equipment procurement cycle targeting advanced process nodes where Applied Materials has historically strong share.

What the Spending Plans Mean for Equipment Makers

Applied Materials' competitive strength lies in thin-film deposition, ion implantation, and chemical mechanical planarisation, processes used heavily in DRAM stack construction and back-end processing for memory devices. As SK Hynix and Samsung scale their HBM stacks from eight to twelve or sixteen-layer configurations to match AI memory bandwidth requirements, the per-wafer equipment intensity of their production increases, translating to higher equipment revenue per unit of capacity added.

Lam Research occupies a complementary position in the same supply chain through its etch and selective removal capabilities, which are critical for high-aspect-ratio structures in advanced DRAM. Lam's exposure to Korean memory capex cycles has historically been proportional to Applied Materials', making it a secondary beneficiary of the same spending announcement.

The AI Memory Demand Connection

The capacity expansion at Samsung and SK Hynix is tied directly to AI infrastructure buildout: the limiting factor in deploying large language model inference at scale is not GPU compute alone but HBM bandwidth. Both companies are therefore responding to customer commitments from hyperscalers and AI chip designers rather than speculative demand. This makes the capex announcement more durable than a cyclical memory recovery, as the customer orders backing the capacity investments are longer-dated commitments.

For equipment makers, the combination of leading-edge DRAM expansion in Korea and ongoing logic expansion by TSMC and Intel creates a backdrop of sustained advanced-node equipment demand that extends through at least the near-term planning horizon of both Applied Materials and Lam Research.