Nvidia, SK to detail cooperation plan as Huang flags prolonged chip shortage - CNA

NVIDIA and South Korea's SK Group are set to unveil a cooperation plan on June 8, 2026, amid ongoing global chip shortages. NVIDIA CEO Jensen Huang emphasized that the current memory chip shortage wil... Read original →

Jensen Huang’s warning of a multi-year memory shortage exposes the structural strain AI compute places on HBM3E/HBM4 and advanced packaging. NVIDIA’s SK Group alliance isn’t just about capacity—it’s a vertical integration play locking in SK Hynix’s HBM and silicon photonics to secure memory-to-interconnect co-design. This pressures Micron to accelerate CoWoS alternatives and pushes TSMC toward deeper collaboration with Samsung in advanced packaging. Geopolitically, the U.S.-ROK axis aims to reduce reliance on Taiwan, China’s outsized role in back-end testing, yet SK Hynix’s DRAM fabs in mainland China remain a compliance vulnerability. Over the next 18 months, HBM yield and photonic integration—not GPU availability—will bottleneck AI supercomputer deployment. Failure to ramp HBM4 by mid-2027 risks systemic delays for NVIDIA’s Blackwell Ultra architecture.