Skip to content
News Funding & M&A

Samsung Group commits $648 billion over ten years to AI chips and data centers in South Korea

· by Pondero Newsdesk

The short version

Samsung Group plans to invest 1,000 trillion won (roughly $648 billion) over the next decade on semiconductor fabs, AI data centers, batteries, and displays. The largest industrial pledge in South Korea's history.

Samsung Group commits $648 billion over ten years to AI chips and data centers in South Korea

Samsung Group announced plans on June 26 to invest 1,000 trillion won (approximately $648 billion) across the next decade, targeting semiconductor fabrication, AI data centers, battery production, and advanced displays in South Korea. A senior Samsung executive presented the package to President Lee Jae Myung at a meeting on June 26, per reporting from SammyGuru citing The Information and Reuters. The figure makes this the largest industrial spending commitment in the country's history.

What

The plan covers four sectors: semiconductor manufacturing, AI data centers, batteries, and display technology. Roughly 300 trillion won of the total is earmarked for new chip fabrication facilities in South Korea's southwest, a region chosen partly because the area around Seoul is running short on land, water, and power capacity needed to operate large-scale fabs. Samsung's existing chip operations are concentrated near the capital.

The geographic shift also aligns with the South Korean government's stated goal of distributing industrial growth beyond the Seoul metropolitan zone. Not all observers welcome the rationale. Per SammyGuru's reporting, some critics argue the government is nudging companies toward politically preferred regions, and others raise concerns that dispersing investment could dilute the efficiency gains of concentrated semiconductor clusters.

SK Hynix executives attended the same June 26 presidential briefing, and analysts expect rival memory chipmakers to follow with comparable domestic commitment packages.

Why it matters

The bottleneck slowing AI training cluster buildouts right now is HBM memory, not model architecture or software. Samsung is one of three manufacturers of high-bandwidth memory globally (alongside SK Hynix and Micron), and demand for HBM from data center operators has outpaced supply since late 2024. A decade-long, $648 billion commitment to expand fabrication in South Korea is a direct attempt to remove that constraint from the supply chain.

For AI-tool operators and enterprise buyers, the implication is practical: more fab capacity over the next few years points toward reduced HBM scarcity, which feeds into GPU availability at hyperscalers and cloud AI pricing. This announcement does not resolve a shortage today, but it is the kind of long-cycle infrastructure bet that shapes what compute costs in 2028 and beyond.

The scale also changes the competitive picture for TSMC and Intel Foundry. A fresh Samsung investment wave aimed at non-Seoul regions, with government backing, signals South Korea intends to defend its position as the primary HBM supplier while also building out a broader AI hardware stack domestically.

Context

Samsung's announcement came the same week demand reports from major AI cloud operators showed memory bottlenecks persisting into the second half of 2026. The Group has faced pressure from analysts to respond to SK Hynix's lead in HBM3E market share, where Hynix has supplied the bulk of Nvidia's H100 and H200 memory to date. Samsung has been working to qualify its own HBM3E chips with Nvidia and is expected to ramp certified supply later this year.

The broader context is a wave of national industrial policy coordinated between tech firms and governments. South Korea's announcement mirrors large domestic investment pledges from TSMC in Japan and Taiwan, Intel in the United States under the CHIPS Act framework, and ASML's facility expansions in the Netherlands.

What to watch next

Watch for Samsung's official announcement of the plan, which per reporting was expected around the June 26 presidential meeting but had not yet been formally published as of this writing. Two specific signals will clarify the plan's credibility: a phased capex schedule showing when the first 300 trillion won in southwest fab construction is expected to begin, and whether Nvidia formally certifies Samsung HBM3E in time to benefit from the next GPU generation ramp.

Sources