Nvidia and SK Group ink $500 billion AI infrastructure partnership

This partnership addresses the critical shortage of high-bandwidth memory (HBM) that currently limits the scaling of AI training clusters. By locking in SK hynix’s HBM4 supply, Nvidia creates a significant barrier for competitors like AMD and custom silicon providers seeking equivalent memory performance. For the streaming and enterprise sectors, the 2-gigawatt deployment signal a shift toward localized 'sovereign AI' infrastructure that prioritizes energy-efficient token generation and geographic data residency. Watch for the 2027 operational milestone as a benchmark for when this massive increase in regional compute capacity begins affecting global spot prices for AI inference.

The $500 billion-plus commitment follows a broader trend of sovereign AI investments aimed at decoupling compute dependence from traditional hyperscalers. According to Reuters in July 2026, the deal was announced during a high-profile meeting between Nvidia CEO Jensen Huang and South Korean President Lee Jae Myung, underscoring the strategic nature of the alliance. The deal is effectively a full-stack integration, combining Nvidia’s software and GPU designs with SK hynix’s and potentially Samsung’s advanced memory manufacturing to ensure supply continuity through the end of the decade.

Simultaneous to the SK Group announcement, Nvidia revealed a $1 billion investment in Naver to expand its Korean data center footprint, per Tom's Hardware reporting in July 2026. This activity is part of a larger $950 billion national AI push by the South Korean government, which includes a separate $200 billion memorandum of understanding between Samsung Electronics and Broadcom for foundry and memory chips through 2030. These moves reflect a calculated effort by Nvidia to diversify its manufacturing and infrastructure base amid heightening trade restrictions and rising energy costs in the United States.

Energy constraints remain a primary hurdle for these gigawatt-scale projects. While SK Telecom’s planned 2-gigawatt facility is unprecedented—dwarfing typical 50-200 megawatt data centers—industry analysts note that South Korea’s power grid and water resources will face significant strain. According to internal documents cited by MLQ.ai in July 2026, SK Telecom has launched a subsidiary called SK Hyper with the long-term goal of reaching 15 gigawatts of domestic capacity by 2035, positioning the firm as a primary anchor for national AI infrastructure alongside partners like Hyundai and Hanwha.