A planned 18-day strike at Samsung Electronics is shifting investor attention from labor negotiations to the risk of memory-output disruption, as estimates from Korean media point to significant potential losses across the company's semiconductor operations.
Micron's senior vice president, Jeremy Werner, told The Circuit Podcast that memory has become a strategic bottleneck for data-center inference, warning that insufficient memory can sharply cut GPU utilization while faster, larger memory can theoretically multiply the compute extracted from GPUs. The remarks underscore how storage and memory design could limit AI deployment.
Samsung Electronics' labor and management have returned to the negotiating table under South Korean government mediation, as the company faces a planned general strike and a widening dispute over how profits from its semiconductor rebound should be shared.
Phison Electronics Corp. reported record revenue and profit for the first quarter of 2026, while Chief Executive Khein-Seng Pua warned that a severe imbalance in the global NAND market would push average selling prices higher and sustain tight supply conditions into the second half of the year. He said the company is accelerating a strategic shift into AI storage infrastructure and edge AI computing as part of its transition, which it calls Phison 3.0.
LG Innotek is seeking to enter Tesla's AI semiconductor supply chain by targeting ABF-based FC-BGA substrate orders for the automaker's AI4 chips, intensifying its competition with Samsung Electro-Mechanics (Semco) across both autonomous driving and humanoid robotics hardware.
AI-driven demand and new product cycles are accelerating competition among major memory makers, creating pressure across storage and component supply chains. Production is running near full capacity, margins on some DDR5 products have improved, and next-generation HBM will be a focal point of contention through 2027 and beyond.
Advanced Semiconductor Engineering (ASE) and PCB manufacturer Wus Printed Circuit are teaming up to build a new packaging facility in southern Taiwan, as the world's largest chip assembler moves to expand capacity ahead of surging AI-driven demand. The plant, to be located in the Kaohsiung Nanzih Technology Industrial Park, is scheduled to open in September 2029.
To secure a foothold in the booming market for advanced AI chip packaging and testing, Taiwan's outsourced semiconductor assembly and test (OSAT) industry is entering what executives and analysts describe as an unprecedented expansion cycle. Industry observers estimate that combined capital expenditures by Taiwanese OSAT providers could surge to a record NT$400 billion (approx. US$12.7 billion) in 2026, driven by an aggressive race to expand high-end production capacity.
Below are the most-read DIGITIMES Asia stories from the week of May 4-10, 2026:
The global copper-clad laminate market expanded sharply as AI computing demand forced higher-spec hardware, and Taiwan's printed circuit board supply chain moved to secure second-source positions, industry data showed. The Taiwan Printed Circuit Board Association reported the copper-clad laminate market reached US$16.02 billion in 2025 and was projected to rise to US$21.5 billion in 2026 as server designs demanded larger sizes, ultra-low-loss performance, and boards with more than 40 layers.
Global memory supply is under severe strain, prompting major cloud service providers to sign multi-year contracts with memory makers and forcing the industry to rethink capacity spending and pricing. The shortage reflects long lead times for new fabs and a strategic shift toward higher-margin memory types.
More coverage
