Samsung and SK Hynix, which together produce about two-thirds of the world's memory chips, announced a $518 billion plan to build four new chip factories in South Korea's southwest region. South Korea's president stood alongside both company chairs at a public briefing, framing the effort as a national mission: semiconductors, AI data centers, and physical AI as the three pillars the country must build faster than anyone else.
To understand why this matters, you need to know what is happening to memory chips right now. The kind of memory chips used in AI data centers requires enormous amounts of production capacity. Data centers building AI infrastructure now consume an estimated 70% of all memory chips produced globally. That share has risen sharply and continues to grow.
The three companies that produce nearly all of the world's memory, Samsung, SK Hynix, and US-based Micron, have responded rationally: they shifted their factories toward making the higher-margin chips that AI companies will pay a premium for. The side effect is that far fewer chips are available for everything else: laptops, smartphones, cars, industrial equipment, and consumer electronics.
The result is a shortage that analysts describe as structural, not cyclical. DRAM prices, the type of memory used in virtually every electronic device, rose sharply through 2025 and have continued rising in 2026. Analysts at IDC and Kearney expect the shortage to persist at least through 2027, with some projections extending to 2030. The Micron CEO said in June 2026 that supply will only gradually improve by 2028.
For business operators in any sector that procures electronics, equipment, or technology, this is a direct cost input story. Higher memory prices are already showing up in laptops, phones, and technology infrastructure. Companies building or refreshing their IT equipment in 2026 and 2027 will pay more than they would have two years ago. This is not speculation: Dell, Honda, and Qualcomm have all stated publicly that memory cost increases are hitting their products and margins.
The new Samsung and SK Hynix factories are a genuine long-term response. But SK Hynix's own chairman was candid: it took nine years to build their current main manufacturing cluster. The companies have not given a completion date for the new facilities. The government promised to speed up permit approvals, and there are questions about whether the chosen southwest region has enough power and water to support this scale of industrial activity.
Investors were not impressed, at least not on day one. Samsung shares fell nearly 5% and SK Hynix fell close to 2% on the announcement, reflecting concern about the scale of capital required, the infrastructure challenges, and the possibility of oversupply further down the road if AI demand ever slows.
The broader picture is a global race to build chip capacity. South Korea is making its move. The US, Taiwan, Japan, and the EU are all running similar programs. For any business that buys electronics, machinery, or technology, this race determines your input costs for the next decade. The factories being announced today are the supply that might, eventually, bring prices back down. But not soon.