Particle.news
Download on the App Store

South Korea Converts Megaprojects into $950 Billion of AI and Chip Partnerships

The deals aim to secure GPU and high‑bandwidth memory supply to drive multi‑gigawatt AI data centers while key sites, financing and infrastructure remain to be settled.

Overview

  • At a San Francisco summit on July 24, President Lee Jae Myung unveiled the “San Francisco AI Declaration” and helped seal MOUs and partnership commitments totaling about $950 billion.
  • The headline commitments include SK Group pursuing roughly $750 billion in long‑term memory supply deals, Samsung signing an up‑to‑$200 billion MOU with Broadcom, and a $500 billion‑plus Nvidia–SK initiative focused on AI data centers and next‑generation memory.
  • SK Telecom plans a 2‑gigawatt AI data center using Nvidia’s Vera Rubin systems and SK Hynix HBM4 memory with the first facility targeted to be online in 2027.
  • Nvidia said it secured long‑term HBM supply from SK Hynix, will co‑develop next‑generation HBM, and will invest $1 billion in Naver to expand GPU‑based data center capacity.
  • Most announcements are MOUs or high‑level commitments and face major execution risks, including finding sites, raising phased financing, meeting huge power and water needs, building a skilled workforce and avoiding future overcapacity.