As artificial intelligence (AI) applications expand globally, United Microelectronics Corporation (UMC) is intensifying its efforts in advanced packaging and silicon photonics (SiPh) technologies. UMC co-president Jason Wang highlighted that although these segments currently generate modest revenues, a surge in project activity is anticipated to drive "significant growth" beginning in 2027, marking a strategic focus on emerging markets tied to AI and related fields.
UMC is leveraging its advanced packaging capabilities to serve a broader range of applications beyond high-end computing, including mobile devices, power management integrated circuits (PMIC), BCD processes, and AI products. Since UMC does not fabricate memory wafers internally, it primarily operates through collaboration models in this area, where customers supply memory wafers, and UMC performs bonding and process integration.
Wang also addressed concerns about substrate layer capacity, clarifying that expansion discussions mainly target post-2027 timeframes, as 2026 will prioritize product launch and volume ramp-up. Some existing equipment used for 40nm and 65nm processes may be adapted to meet demand, with further investment decisions hinging on customer project visibility and long-term requirements.
UMC expects advanced packaging revenues, currently sourced from a limited customer base, to gain momentum as it adds more than 20 projects in 2026 alongside the over 10 partnerships already established. This growth trajectory aligns with broader industry trends underscoring AI-driven semiconductor demand.
UMC's path to silicon photonics commercialization
UMC's strategic push into SiPh includes a technology licensing agreement with Belgium's Imec to acquire the Imec iSiPP300 process. This initiative aims to expedite the development of a 12-inch SiPh platform capable of producing photonic chips for optical transceivers. The company is working closely with multiple new customers to drive risk production runs planned for 2026 and 2027, highlighting a phased approach toward commercialization.
Industry-wide, Wang noted that mature process supply is increasingly constrained, a dynamic he attributes to "structural" crowding-out by AI applications rather than short-term disruptions such as those experienced during the COVID-19 pandemic. He added that the long-term nature of AI demand makes large-scale investments in mature process capacity economically unfeasible. Consequently, tight supply-demand balances may persist, though the current phase remains early in the evolution.
On pricing, Wang observed an improvement in overall foundry prices but emphasized that high utilization alone does not elevate average selling prices (ASP). Pricing decisions must consider long-term competitiveness for both UMC and its customers. Gross margins are under ongoing pressure from higher depreciation, inflation, and exchange rate fluctuations. UMC's first-quarter lower gross margin primarily reflected higher depreciation expenses, with cost challenges expected to continue throughout 2026.
The company's outlook signals cautious optimism rooted in technological advancements and expanding workloads driven by AI, balanced by market uncertainties in capacity and cost management.
Article translated by Jingyue Hsiao and edited by Jack Wu
