Samsung and SK Hynix Plan $590 Billion Chip Investment as AI Demand Drives Memory Prices Higher
Samsung and SK Hynix, the world’s two largest memory chip makers, are jointly planning a combined investment of roughly $590 billion in South Korea’s semiconductor industry over the coming years. The massive capital injection is a direct response to soaring demand for high-bandwidth memory (HBM) chips used in AI data centers, which has sent memory prices skyrocketing.
South Korea’s government will support the initiative with tax breaks and infrastructure funding, aiming to build a mega chip cluster near Seoul. The investment is expected to create over 100,000 new jobs and solidify the country’s position as the global leader in memory manufacturing.
The Lede: Why This Investment Matters Now
AI workloads require immense amounts of fast, power-hungry memory. HBM chips are the critical bottleneck in training large language models and running inference at scale.
Memory prices have risen sharply since mid-2023, with HBM selling at a premium. Samsung and SK Hynix together control more than 90% of the global HBM market. This investment is designed to secure that dominance and meet exploding demand.
Breakdown of the Investment Plan
- Samsung Electronics expects to spend about $230 billion on new fabs, R&D, and packaging facilities over the next 20 years. The company is already ramping production of HBM3E chips for Nvidia’s next-generation AI accelerators.
- SK Hynix plans to invest roughly $180 billion in its own capacity expansion, focusing on advanced HBM and next-generation DRAM. The firm recently announced it has secured a multi-year supply agreement with Nvidia.
- South Korea’s government will contribute approximately $180 billion in tax incentives, power supply subsidies, and land grants. The aim is to create a massive “K-Semiconductor Belt” spanning multiple cities.
“This is not just a business decision — it’s a national security imperative. AI chips will define the next decade of computing, and South Korea intends to lead.”
— South Korean Minister of Trade, Industry and Energy (paraphrased from The Decoder’s source)
How AI Demand Is Reshaping the Memory Market
AI systems need far more memory than traditional computing. A single Nvidia H100 GPU uses 80GB of HBM3 memory. The upcoming B200 “Blackwell” GPU will need up to 192GB per module.
This has created a structural shift. HBM supply is now the limiting factor in AI server deployment. Prices for HBM have risen more than 300% year-over-year, according to industry analysts.
Samsung and SK Hynix are both racing to develop HBM4, the next generation, which will offer even higher bandwidth and lower power consumption. The $590 billion investment plan aims to have HBM4 in mass production by 2026.
What This Means for the Chip Industry
The investment will likely intensify competition with Micron, the third-largest memory maker, which is based in the U.S. Micron has also announced plans to build new HBM fabs in Idaho and New York.
But the scale of South Korea’s commitment is unmatched. If fully executed, the combined investment would be the largest single industrial project in the country’s history.
Analysts warn that supply eventually may outpace demand, especially if the AI boom slows. However, both companies and the government are betting that AI adoption will continue to accelerate for at least the next decade.
Potential Risks and Challenges
- Geopolitical tensions with China and Taiwan could disrupt supply chains. South Korea relies on imported machinery and materials.
- Workforce shortages in chip engineering and packaging are already acute. The plan assumes the country can train enough skilled workers.
- Environmental impact from water and energy use in fabs is a growing concern. Local communities have protested new factory construction.
The Bigger Picture: A Global Chip Race
The U.S. CHIPS Act, the European Chips Act, and Japan’s domestic subsidies are all driving a global race to build AI-capable chip capacity. South Korea’s $590 billion plan is the boldest response yet.
Memory prices will likely remain elevated through 2025, benefiting both Samsung and SK Hynix’s bottom lines. But the true test will be whether the investment translates into sustained technological leadership as AI models grow ever larger.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.