Author：Wall Street CN

According to DigiTimes, citing industry sources, Micron plans to invest $24 billion in Singapore to expand its NAND flash memory production capacity. This project will require 400 to 500 power transformers, more than double the 100 to 150 typically needed in a standard wafer fab. This scale exceeds the annual production capacity of any transformer manufacturer in Taiwan, highlighting heavy-duty power equipment as a major bottleneck for AI-driven semiconductor capacity expansion.

Micron's enormous demand reflects the extremely high power intensity of next-generation memory chip factories related to AI. Tight production capacity of high-bandwidth memory (HBM) for AI servers has driven major memory chip manufacturers to expand production simultaneously, but the power infrastructure supply required to support these factories is failing to keep pace with the increasing demand.

Micron's Singapore project, scheduled to begin production by the end of 2028, is part of its large-scale global expansion plan. The company has acquired Powerchip Technology's Miaoli Tongluo plant in Taiwan for $1.8 billion, which is expected to be operational in 2026; it is also building new plants in Idaho and New York in the United States, and its plant in Hiroshima, Japan, is expected to begin production in the second half of 2026.

Samsung Electronics and SK Hynix have also announced expansion plans, both driven by the same demand logic: the consumption of HBM (Heavy Machine Tool) by AI servers far exceeds the supply capacity of existing production lines. This has sparked a wave of simultaneous factory construction across three continents, with various projects vying for the same resources of heavy electrical equipment and raw materials.

This impact is already reflected in prices and supply: Due to a surge in orders and rising costs of raw materials such as copper, major Taiwanese heavy electrical equipment suppliers such as Fuji Electric and Chemi-Tech Electronics have raised prices by 20% to 30%. Meanwhile, some transformer manufacturers, unable to meet stringent delivery cycles and large-volume demands, have refused to quote for large semiconductor projects. Industry insiders say that currently no single manufacturer can independently handle large-scale orders from the AI and semiconductor industries.

International transformer brands are increasing their market share despite higher prices, thanks to their larger overseas factories with greater production capacity. Taiwanese manufacturers, on the other hand, are beginning to collaborate with second-tier suppliers, breaking down specifications and production capacity to jointly meet customer needs.

Transformers are general-purpose critical equipment, required not only by wafer fabs but also by AI data center construction, large-scale energy storage projects, and power grid expansion. The already strained supply chain before the AI production boom has now become a battleground for hundreds of units per project.

Delays in transformer delivery will likely lead to delays in wafer fab production, which in turn will hinder the mass production of memory chips that AI companies rely on. Operators planning to build new data centers are also in the same boat, competing with semiconductor companies for this type of equipment with production and delivery cycles that can take months.

This article is sourced from: