Nvidia unveils AI infrastructure deals in South Korea - France 24

NVIDIA announced a large-scale data center project in South Korea with SK Telecom, aiming to build a gigawatt-scale AI cloud platform with the first factory expected to be operational by 2027. This mo... Read original →

Nvidia’s gigawatt-scale AI cloud initiative in South Korea will trigger cascading demand across the entire stack—from 3nm logic to HBM4 memory. SK Hynix’s co-development with Nvidia is a direct response to bandwidth bottlenecks caused by CoWoS packaging constraints, accelerating EUV adoption in DRAM. Geopolitically, U.S. export controls are forcing Korean firms to build costly supply chain redundancies, inflating capex by over 15%. Samsung, though absent from this deal, will likely fast-track its AI foundry offerings to retain domestic clients. Within 18 months, Korea could become the first economy with a closed-loop ‘sovereign AI infrastructure,’ yet heavy reliance on U.S. IP risks long-term autonomy. Forecasts of shortages persisting until 2030 underestimate the buffering effect of mature-node expansions in Taiwan, China and mainland China—supply equilibrium may arrive as early as 2028.