The AI-chip supply chain is drawing investor attention beyond the companies designing and selling accelerators. Three names highlighted in a recent market screen—ACM Research, South Korea’s PSK and Japan’s Sumco—occupy different positions in the manufacturing chain, leaving each exposed to a different balance between demand, capacity and pricing.
ACM Research makes semiconductor capital equipment used for wafer cleaning and advanced packaging. Its exposure is tied to the increasing complexity of AI chips, including larger dies and 2. The company was also described as booking production and evaluation orders for panel-level electroplating, while advanced-packaging revenue was reported to be growing at a triple-digit rate.
PSK supplies semiconductor-processing equipment, including dry-strip and 3D integrated-circuit packaging tools. The company’s position depends on foundries continuing to invest in capacity for AI and high-performance computing, making future fabrication spending an important variable for its results and valuation.
Sumco operates further upstream by supplying high-purity silicon wafers used to manufacture advanced chips.
That exposure sits against a supply chain in which the tightest constraints have not necessarily been the logic dies themselves. An independent supply analysis estimated that Nvidia, Google, AMD and Amazon together consumed more than 90% of global CoWoS advanced-packaging capacity and HBM supply by value in 2025, compared with about 12% of advanced logic-die production.
Nvidia’s results show why that bottleneck matters to the broader chain. Their performance will therefore depend not only on continued AI demand, but also on whether capacity expansion preserves pricing power in the specific layer each company serves.