Artificial Intelligence (AI) for Semiconductor Packaging

Journal Article (2026)
Author(s)

Cadmus Yuan (Feng Chia University)

Xuejun Fan (Lamar University)

Xinyu Liu (Lamar University)

G.Q. Zhang (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Research Group
Electronic Components, Technology and Materials
DOI related publication
https://doi.org/10.1109/ACCESS.2026.3740841 Final published version
More Info
expand_more
Publication Year
2026
Language
English
Research Group
Electronic Components, Technology and Materials
Journal title
IEEE Access
Volume number
14
Pages (from-to)
155421-155432
Page Views
12
Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

Artificial intelligence is increasingly being applied to semiconductor packaging design, manufacturing, and reliability. However, most existing models remain isolated and problem-specific surrogates trained on small and narrowly defined datasets with limited numbers of inputs and outputs. This perspective presents a strategic roadmap toward scalable, transferable, and adaptive intelligence for semiconductor packaging. We identify the scarcity, fragmentation, and inconsistent representation of diverse data as the central barriers to AI adoption in semiconductor packaging. To address these challenges, we call for a coordinated global effort to establish shared data infrastructure with standardized representations, comprehensive metadata, traceable provenance, and quantified uncertainty. We propose simulation as a data factory, using consistent multi-fidelity simulations complemented by experimental measurements, manufacturing observations, and field data to systematically expand the scale and coverage of semiconductor packaging datasets. Building on this infrastructure, we envision a pre-trained, modular, and extensible foundation model that learns transferable representations and adapts to downstream tasks across co-design, manufacturing, and reliability. Ultimately, this roadmap seeks to advance semiconductor packaging AI toward transferable, scalable, adaptive, and continuously validated engineering intelligence.