Industry Analysis
Huang’s meeting with Chey signals a strategic pivot ahead of HBM3E/HBM4 mass production. Technically, NVIDIA is offloading its AI memory-wall pressure onto SK hynix—stacking beyond 12 dies and TSV pitch below 3μm will force material suppliers to adopt low-alpha substrates and advanced thermal interface materials. On compliance, tighter U.S.-ROK export controls mean SK’s China-based HBM expansion risks triggering U.S. equipment re-licensing, inflating capex by over 15%. Samsung will likely accelerate GDDR7 lobbying to undercut mid-tier AI training demand, while Micron leverages Intel MI300X integration to promote North American supply resilience. Within 18 months, HBM will shift from a performance differentiator to a mandatory AI chip requirement, institutionalizing joint CoWoS-HBM capacity reservations between TSMC and SK hynix.
This page displays AI-generated summaries and metadata for research purposes. Original content belongs to the respective publishers.