AI Infrastructure Race: Nvidia Dominates as Korean Chipmakers Face Investment Risks
Global tech giants are pouring unprecedented investment into artificial intelligence infrastructure, with a combined $650 billion (approximately ₩941 trillion) slated for deployment this year. However, the benefits of this massive spending are heavily skewed towards US-based Nvidia, leaving Korean semiconductor companies vulnerable to potential downturns if AI investments fail to deliver expected returns.
Nvidia’s Dominance and the AI Investment Boom
Nvidia currently controls an estimated 86-90% of the AI accelerator market, solidifying its position as the primary beneficiary of the current investment cycle. The company reported a 75.2% adjusted gross profit margin in its most recent quarterly report, demonstrating the profitability of its AI-focused products. Nvidia News
This surge in investment isn’t limited to chip manufacturing. Companies like Oracle have committed to substantial long-term contracts, including a $300 billion deal announced in September 2025 to supply large-scale computing power. Nvidia News OpenAI is as well diversifying its cloud infrastructure providers, initially exploring a $500 billion data center project dubbed “Stargate” with Oracle and SoftBank, though implementation has faced delays due to leadership competition and funding issues. Nvidia News
Korean Semiconductor Industry at a Crossroads
Samsung Electronics and SK Hynix are key suppliers of High Bandwidth Memory (HBM), a critical component for AI data centers and Nvidia GPUs. Together, they hold over 90% of the global HBM market share. Nvidia News While this positions them to benefit from increased AI investment, they are also uniquely exposed to risks if the investment cycle slows or profitability falters.
Analysts predict a combined operating profit of ₩375 trillion for Samsung and SK Hynix in 2026, contingent on continued positive returns on AI investments. CNBC However, a decline in big tech profitability could quickly evaporate these forecasts, leading to order cuts and reduced revenue for Korean chipmakers.
HBM Supply and Demand Dynamics
The HBM market is currently experiencing strong demand, driven by Nvidia’s needs. However, concerns about potential oversupply are emerging, particularly regarding the HBM3E generation. Some investment banks, including Morgan Stanley, warn that aggressive capacity expansion by Samsung, SK Hynix and Micron could lead to price declines in the second half of 2026. CNBC
However, other market research agencies, such as Trend Force and IDC, anticipate continued strong demand for Nvidia’s next-generation accelerators, suggesting that HBM4 supply will remain constrained until the first half of 2027. CNBC The time required to stabilize HBM4 mass production yield and the expanding inference server market are also expected to support continued demand.
SK Hynix’s Ascendancy and Samsung’s Response
In 2025, SK Hynix surpassed Samsung Electronics in operating profit for the first time, largely due to its dominance in the HBM market and its focus on memory chips. CNBC SK Hynix posted a record operating profit of ₩47.2 trillion, compared to Samsung’s ₩43.6 trillion. CNBC
Samsung is actively challenging SK Hynix’s lead, recently announcing the commercial shipment of HBM4 chips to customers, including Nvidia. Bloomberg The competition between the two Korean giants, along with Micron Technology, is intensifying as they vie for Nvidia’s lucrative HBM orders.
Northeast Asia’s AI Imperative
Despite short-term risks, experts believe that AI demand in Northeast Asia will remain structurally strong. Facing demographic challenges and labor shortages, countries like Korea and Japan are increasingly reliant on AI and robotics for economic survival. Nvidia News
Key Variables to Watch
Two key factors will shape the future of the AI infrastructure market in 2026: whether big tech companies can generate real revenue from AI services and whether the mass production of HBM4 can be successfully scaled to address potential oversupply concerns. Nvidia News