HBM4

NVIDIA and Samsung are deepening a partnership that now reaches far beyond routine supplier ties. Their latest talks in Seoul point to a longer-term strategy built around advanced semiconductors, high-bandwidth memory, and the next wave of AI infrastructure. The most important signal from the meeting is not just what the two companies are working on today, but how clearly they are positioning themselves for future AI platforms. That includes HBM4 and HBM5, along with broader chip collaboration that could shape both computing and manufacturing priorities. HBM Becomes A Strategic Battleground High-bandwidth memory has become essential to modern AI accelerators because it directly affects data transfer speed and processing efficiency. For NVIDIA and Samsung, that makes memory supply a core part of the competition around future AI systems. Jensen Huang confirmed that Samsung has now been certified to supply memory chips for NVIDIA’s Vera Rubin platform. For that platform, Samsung will provide HBM4, a significant step that strengthens its position in the AI memory market. Samsung has also shared samples of HBM4E, an upgraded version designed to keep pace with rivals such as SK Hynix and Micron. The move underscores how intense the competition has become in this segment. Work Extends Into Automotive And AI Inference The collaboration between the two companies is not limited to memory. Jun Young-hyun said Samsung and NVIDIA are also working on chips for self-driving vehicles, showing that the relationship now spans multiple high-growth categories. They are also involved in development tied to the AI inference accelerator Groq. NVIDIA previously confirmed that the LP30 Groq AI inference processor will be manufactured by Samsung, with shipments scheduled for the end of this year. That mix of automotive, inference, and memory work shows how Samsung is becoming more deeply embedded in NVIDIA’s broader AI ecosystem. It also suggests that the partnership is being built for scale, not for a single product cycle. Leadership Talks Signal A Broader Long Game Huang was in South Korea this week to meet local partners, including leaders in the semiconductor industry. One of his main meetings was with Jun Young-hyun, Samsung Electronics’ co-CEO and head of its chip division. Talks reportedly covered not only current projects but also next-generation semiconductors in a wider sense. The inclusion of HBM4 and HBM5 in those discussions shows that both sides are looking well beyond immediate supply needs. A meeting with Chairman Jay Y. Lee was considered unlikely during this trip, as he was reported to be outside South Korea on business. Even so, top-level contact between the two companies has remained frequent in recent months. That consistency matters because both sides have strong reasons to keep the relationship close. NVIDIA needs partners that can support AI chip production, advanced memory, and future platform requirements, while Samsung needs a stronger foothold in the global AI supply chain. The result is a partnership that now carries strategic weight across multiple fronts. From Vera Rubin and HBM4 to HBM5 planning and manufacturing for LP30 Groq, the Seoul meetings make clear that the next phase of AI competition will depend as much on memory and production capacity as on chip design. Source: www.sammobile.com

NVIDIA And Samsung Tighten Their AI Chip Bet, HBM4 And HBM5 Move To The Center