AMD’s chief executive Lisa Su is due in Seoul next week to meet Samsung Electronics chairman Lee Jae‑yong in discussions aimed at securing high‑bandwidth memory (HBM) for the company’s accelerating artificial‑intelligence chip programmes. The talks, which also include a prospective meeting with South Korea’s largest internet portal, Naver, underscore AMD’s push to shore up component supply and to explore partnerships around AI compute infrastructure in the region.
HBM is a specialised form of memory used alongside modern AI accelerators; it sits close to the processor and delivers vastly higher bandwidth than traditional DRAM. Demand for HBM has surged with the proliferation of generative‑AI workloads, yet supply remains thin relative to appetite, creating a bottleneck that can determine which chipmakers can scale their data‑centre products quickly and cost‑effectively.
For AMD the objective is straightforward: secure a predictable flow of HBM so its GPUs and custom accelerators can compete for large cloud and enterprise orders without being hamstrung by component shortages. Samsung is one of a handful of manufacturers that can produce advanced HBM, and a closer commercial relationship would give AMD an important hedge against volatile spot markets and allocation battles with rivals.
The outreach to Naver signals a broader strategy than just component procurement. By discussing AI compute infrastructure with a major regional cloud and internet operator, AMD is positioning itself to tie silicon deals to local deployments of AI systems — a model that promises faster time to revenue and tighter customer integration, while deepening ties with Korean technology players.
The move comes amid a frenzied global competition for the feedstock of AI: not only chips but the memory, packaging and systems that together make large models run. Nvidia’s dominance in AI accelerators has intensified the scramble among suppliers and customers for reliable inputs; AMD’s approach of bilateral industry deals is a pragmatic response to that pressure and to the broader need for supply‑chain resilience.
If successful, the talks could accelerate Samsung’s role as a strategic memory partner for hyperscalers and AI‑chip vendors beyond its traditional customer base. For Korea’s technology sector, closer ties with a major US chip designer may bring commercial benefits and bolster the region’s importance in the global AI hardware ecosystem, even as geopolitics and trade policy continue to reshape semiconductor sourcing decisions.