Nvidia CEO Jensen Huang's bold moves on the show floor at the annual developer conference "GTC 2026" in San Jose, California, on the 16th (local time) have become the talk of the industry. Shouting "Samsung is the best" at the Samsung Electronics booth and drawing a heart while saying "perfect" at the SK hynix booth sent a message that went beyond a simple event. Outwardly it was praise for partner companies, but in essence it is read as a strategic move to hold on to control of the artificial intelligence (AI) semiconductor supply chain.
Behind this "public courtship" lies a structural shift in which decision-making power over AI performance has moved from compute chips (GPUs) to memory. As graphics processing unit (GPU) performance has evened out at a higher level, securing high-bandwidth memory (HBM) to resolve data bottlenecks has emerged as the key variable. The center of gravity in supply chain power is increasingly shifting to memory companies.
Huang's "hands-on management" is seen as an extension of a carefully calculated strategy. From last year's "chicken gathering" with Samsung Electronics Chairman Lee Jae-yong and Hyundai Motor Group Chairman Chung Eui-sun to this booth tour, at the core lies an intent to respond to big tech's moves to reduce reliance on Nvidia (ASIC development). As major customers such as Google, Amazon and Meta speed up in-house chip development, the strategy is read as aiming to lock down the supply chain for HBM, the key component, even if design becomes decentralized.
Amid this trend, Nvidia rival AMD is also moving quickly. AMD CEO Lisa Su visited Korea on the 18th and met with Naver and Samsung Electronics in succession to discuss cooperation. With HBM emerging as the key variable in the AI Semiconductor market, a clear trend is forming of major corporations moving to secure supply chains centered on Korea's memory and manufacturing ecosystem.
In particular, this visit made Nvidia's "dual strategy" toward Samsung Electronics and SK hynix stand out clearly. SK hynix is a key partner that currently has the highest Production yield and technological reliability within Nvidia's supply chain. Huang's affectionate gesture, calling it "perfect," is interpreted as an intent to prioritize securing supply stability at this point in time.
By contrast, the message to Samsung Electronics carried both "expansion and hedging." A case in point was Huang signing not only HBM4 (6th-generation HBM) but also the wafer for Samsung's foundry-made AI chip "Groq." This signals an expansion of cooperation beyond memory into manufacturing and highlights Samsung as an alternative that can ease TSMC's advanced packaging (CoWoS) bottleneck. It is a strategy to maintain supply chain leadership by leveraging Samsung's turnkey capability that spans memory and foundry.
In the end, Huang's public remarks are seen not as mere lip service but as a strategic choice to diversify the supply chain and preserve bargaining power. It reflects an intent to maintain balance by leveraging a triangular structure linking Samsung Electronics, SK hynix and TSMC.
GTC 2026 is assessed as a case showing that the center of gravity in AI Semiconductor competition is shifting to "dominating the supply chain." Around Nvidia, the coupling between memory and foundry is becoming more solid. Industry watchers say even Nvidia has entered a phase where it will be hard to maintain growth speed without Korea's memory and manufacturing strengths.