Why Korea Became the Hottest AI Battlefield in October 2024

HBM: High Bandwidth Memory

High Bandwidth Memory (HBM) stacks multiple DRAM dies vertically using TSV technology, delivering data transfer speeds dozens of times faster than traditional memory. AI model training and inference are limited by data exchange speed, not compute, creating a "memory wall". The latest HBM3e offers up to 1.2 TB/s bandwidth, over ten times DDR5. Nvidia H100 GPUs use HBM3 with 80 GB per GPU.

Global HBM production is dominated by three companies: SK Hynix (50%), Samsung (30%), and Micron (10%). Together, Samsung and SK Hynix control about 80% of the market, and with advanced packaging (3D stacking, PIM) they hold near‑90% of the technology lead.

South Korea accounts for 73% of global DRAM and 51% of NAND shipments, with semiconductor contributing nearly 20% of GDP and generating $141.9 billion in exports in 2024, a 43.9% YoY increase.

OpenAI’s Economic Blueprint

On October 1 Sam Altman flew to Seoul, met President Lee Jae‑myung, Samsung chairman Lee Jae‑yong, and SK Hynix CEO Tae‑won Choi, and signed the Stargate project agreement for two 20 MW data centers and a monthly supply of 900,000 DRAM wafers—more than double current global HBM capacity.

The deal, valued at over $72 billion, lifted Samsung and SK Hynix shares dramatically and pushed the KOSPI above 4,000 points.

OpenAI’s 48‑page "Korea AI Economic Blueprint" (released October 23) proposes a dual‑track strategy: build sovereign AI capabilities (models, infrastructure, data governance, GPU supply) and cooperate deeply with firms like OpenAI to acquire cutting‑edge technology.

The blueprint forecasts AI could raise total factor productivity by 3.2%, adding up to 12.6% potential GDP growth. It predicts AI‑driven medical market growth from $0.377 billion in 2023 to $6.67 billion by 2030 (CAGR 50.8%), AI‑enabled factories from $154.9 billion to $726.4 billion, and a 24% rise in AI‑driven memory chip market.

The Korean government pledged $115 billion in public‑private AI investment, while OpenAI opened its Seoul office (May 2025) as its third Asian location.

Anthropic’s Timeline

Anthropic announced a Seoul office on October 24, following OpenAI’s blueprint release. This is Anthropic’s third Asia‑Pacific office after Tokyo (June) and Bangalore (October 7). Anthropic’s chief business officer Paul Smith highlighted Korean firms as the world’s most mature Claude users, especially for complex programming and enterprise applications.

SK Telecom invested $100 million in Anthropic (August 2023) and co‑developed a multilingual large model supporting Korean, English, German, Japanese, Arabic, and Spanish. The partnership boosted SK Telecom’s customer‑service quality by 34% and enabled Korean startups like Law&Company to double lawyer efficiency with a Claude‑based AI assistant.

Trump’s Deal: $350 Billion Investment for Tariffs

On October 18 Donald Trump played seven hours of golf in Palm Beach with five Korean chaebol leaders (Samsung, SK Hynix, Hyundai, LG, Hanwha). The meeting, arranged by SoftBank’s Masayoshi Son, discussed a $350 billion U.S. investment in exchange for reducing tariffs from 25% to 15%.

The core framework promised Korean investment of $350 billion and a tariff cut, with details to be finalized during Trump’s October 29‑30 APEC visit.

During the APEC summit, Samsung announced a $37 billion chip fab in Texas, SK Hynix a $3.9 billion HBM packaging plant in Indiana, and Hyundai a $26 billion U.S. auto plant, while the U.S. threatened to revoke Chinese‑origin technology exemptions for Samsung and SK Hynix.

Korea’s Strength: Policy, Investment and R&D

Korea aims to be a top‑three global AI power by 2027, with a sovereign AI program backed by $115 billion public‑private funds, a 2025 AI budget of $1.28 billion, and corporate commitments of $48.9 billion by 2027.

The K‑semiconductor strategy targets $450 billion investment by 2030 to secure the global supply chain. In December 2024 Korea passed the AI Basic Act, establishing a national AI committee, safety research institute, and compute center, and imposing transparency obligations on high‑impact AI systems.

Research output is strong: Korea ranks 7th worldwide in AI vitality (Stanford index), leads in AI patent density (17.3 patents per 100 k people), and KAIST ranks 5th globally for top AI conference papers.

Korea’s Weaknesses: Talent, Market and Dependence

Talent retention is poor—Korea ranks 35th out of 38 OECD countries, with only 2,551 AI experts (0.5% of global talent). High‑skill visas to the U.S. and low domestic salaries drive brain drain.

Market size limits growth: Korea’s AI market is $3.12 billion in 2024, far smaller than the U.S. ($150 billion) and China ($80 billion). Export‑oriented economics force Korean AI firms to rely on foreign markets.

Supply‑chain dependence on China for rare‑earths (47.5%) and key materials (70%) poses strategic risk, prompting a goal to cut Chinese imports to 50% by 2030.

Conclusion

Within a single month, Korea witnessed OpenAI’s $500 billion agreement, Anthropic’s office launch, and Trump’s $350 billion investment talks—illustrating how commercial contracts, political endorsement, and technical leadership intertwine.

Korea’s advantage lies in its HBM monopoly, yet it faces talent outflow, a thin AI ecosystem, and regulatory uncertainty that could slow innovation.