As AI-driven semiconductor bottlenecks intensify, global chipmakers are rushing to expand production capacity — accelerating not only the construction of additional fabs but also the planned start dates for facilities already in the pipeline. Even so, the industry widely believes supply will remain insufficient to meet the explosive demand from major technology companies.
Samsung Electronics, the world's top memory chipmaker, has moved up the target operating date for the first of six planned fabs at the Yongin national industrial complex by one to two years — to 2029 from the original 2030–2031 timeline — according to industry sources Monday. The move aims to secure production capacity early to meet surging demand for next-generation AI semiconductors. The complex was originally conceived primarily as a system semiconductor (foundry) site, but there is now growing talk it could be repositioned as a combined memory-and-foundry complex in response to the current memory supply shortage.
Trade, Industry and Energy Minister Kim Jung-kwan had earlier announced at a public briefing on three major megaprojects held at Cheong Wa Dae on June 29 that completion timelines for the final fabs at the Samsung Electronics and SK Hynix semiconductor clusters in Yongin would be significantly advanced. Under the revised plan, Samsung Electronics' Yongin national industrial complex — originally slated for completion in 2047 — is now targeted for 2040, a seven-year acceleration. SK Hynix's Yongin general industrial complex has had its completion date pulled forward by 12 years, from 2045 to 2033.
Building a single semiconductor fab typically takes about two years. For the first Yongin fab to begin operations in 2029, site preparation must begin no later than the second half of this year, with groundbreaking and full structural construction to follow in 2027. Critical infrastructure — including power and water supply — must also be put in place ahead of schedule.
Construction plans at Samsung Electronics' Pyeongtaek campus are also being accelerated. The P4 fab is now on track to reach full operation within this year, roughly six months ahead of the original schedule. The P5 Fab 1, which had been put on hold due to a downturn in the semiconductor industry, resumed full construction in April. Groundbreaking for P5 Fab 2 has been moved up by six months and is set to begin in earnest this month. P5 Fab 1 is expected to come online in 2028, followed by P5 Fab 2 in 2029. Samsung Electronics' memory production capacity is projected to more than double, from 650,000 to 1.35 million 300-millimeter wafers per month. Total investment to build fabs in Yongin, Pyeongtaek and other locations is expected to reach 2,030 trillion won ($1.35 trillion).
The aggressive capacity expansion reflects what the industry broadly sees as a semiconductor supercycle unlike any before. Unlike previous cycles driven by smartphones or PCs, the current surge centers on the buildout of AI data centers, with demand from major technology companies overwhelming available supply.
Analysts also note a structural shift: unlike past expansions marked by high volatility, the current wave of capacity additions is underpinned by long-term supply agreements with global technology companies, providing a stable demand base. Samsung Electronics Chairman Lee Jae-yong has underscored the urgency at a series of recent investment announcements, saying, "This is a moment of reckoning when the foundations of the global economy are shifting."
Micron, the world's third-largest DRAM and HBM (high-bandwidth memory) maker and a close pursuer of Samsung Electronics and SK Hynix, has also been steadily raising its investment commitments. On Thursday, Micron announced plans to expand its US investment to $250 billion by 2035 and to manufacture 40 percent of its DRAM output domestically. The company had originally planned to invest $170 billion in the United States, raised that figure to $200 billion last June following the start of the Donald Trump administration, and has now added another $50 billion.
Separately, Micron has decided to build a next-generation DRAM and HBM factory in Hiroshima, Japan, targeting the second half of 2028 for operations, and held a groundbreaking ceremony Saturday. The investment is valued at 1.5 trillion yen ($9.26 billion).
TSMC, the world's leading foundry, has also said it will raise its US investment commitment from $100 billion to $165 billion to build six wafer fabrication plants, two packaging facilities and one research and development center in the United States.
Behind the three major memory makers' push to front-load investment is also a strategic effort to check the rapid rise of Chinese chipmakers. CXMT, China's largest DRAM manufacturer, plans to raise 29.5 billion yuan ($4.35 billion) through an initial public offering in the middle of this month. The company intends to allocate 7.5 billion yuan toward expanding production capacity, 13 billion yuan toward upgrading its DRAM technology, and 9 billion yuan toward next-generation research and development in a bid to reshape the competitive landscape.
CXMT is also expanding capacity at its existing fab in Hefei and has broken ground on a new DRAM factory in Shanghai. Once the new facility reaches full operation, CXMT's monthly production capacity could grow to as many as 600,000 wafers. US semiconductor research firm SemiAnalysis projects that CXMT's capacity will reach 500,000 wafers per month by 2028, surpassing Micron to become the world's third-largest memory chipmaker.
go@heraldcorp.com