AI 반도체 패권 전쟁의 서막, 젠슨 황의 방한이 남긴 삼성과 인텔의 미래 전략 > K-wave Trends

The most notable aspect of Jensen Huang's visit was his intent to elevate the level of cooperation with Samsung Electronics. Vice Chairman Young-hyun Jun suggested that the meeting was the most productive and positive dialogue in their history, emphasizing a shift from being a simple component supplier to a mid-to-long-term joint development partner. Currently, Samsung is focusing on narrowing the technology gap by sharing its roadmap—ranging from HBM4 and HBM4E to the next-generation HBM5—with NVIDIA. In particular, since HBM4 requires a 4nm foundry process for the logic die, Samsung's "one-stop solution," which combines its memory and foundry capabilities, is expected to play a pivotal role in NVIDIA's next-generation AI accelerator, the "Vera Rubin" platform. This is interpreted as a strategic move to go beyond mere market share competition with SK Hynix and instead embed Samsung’s unrivaled process technology into the customized AI semiconductor ecosystem that NVIDIA desires.

In the foundry sector, moves by global big tech companies to crack TSMC's monopoly are becoming visible. As the bottleneck caused by TSMC's production capacity (CAPA) failing to meet AI semiconductor demand intensifies, Google is betting its survival on supply chain diversification, such as exploring next-generation TPU production through Intel Foundry. Apple has also decided to entrust some chip production to Intel, signaling a clear trend toward dispersing production volume previously concentrated on TSMC. Amidst this, while market expectations are rising—evidenced by Intel's surging stock price—Samsung Electronics' position as an alternative production hub capable of handling both cutting-edge micro-processing and advanced packaging technology is becoming increasingly important. Samsung is steadily expanding its scope of cooperation for next-generation product lines while stably producing NVIDIA’s autonomous driving chips and Groq’s LPUs.

The cards Samsung Electronics is playing to solidify its partnership with NVIDIA are "technological superiority" and "process scalability." Vice Chairman Young-hyun Jun expressed confidence that Samsung's competitiveness will shine in the production of next-generation AI accelerator chips with more complex structures, citing the company's track record in producing autonomous driving chips using 4nm and 8nm processes. In particular, the fact that Samsung Foundry is responsible for producing the "Groq LP30," an inference chip based on technology from Groq (which was acquired by NVIDIA), is a key case proving that NVIDIA's production strategy, previously centered solely on TSMC, is undergoing a real change. By bundling its HBM technology in the memory sector with the advanced process technology of its foundry business through packaging, Samsung is taking a strategy to provide the optimal solution for resolving the complex computational data bottlenecks required by NVIDIA.

Of course, the fact that NVIDIA still identifies SK Hynix as its primary memory partner remains both a challenge and a motivation for Samsung Electronics. Jensen Huang reaffirmed his solid alliance with SK Hynix, expressing his stance that building a stable supply chain across the entire AI ecosystem is his top priority. However, rather than being swayed by this, Samsung Electronics has chosen a direct approach: securing market trust through technological achievements. Vice Chairman Young-hyun Jun's confidence, stating he would "show it through results," is an expression of a strong will to go beyond simply increasing supply volume and to prove mass-production competitiveness by shipping samples of next-generation product lines like HBM4E and HBM5 for the first time in the industry. These efforts by Samsung will serve as a key business driver to regain leadership in the global memory market while becoming the best partner to help NVIDIA succeed.

Ultimately, Jensen Huang's visit served as an opportunity to confirm the "opportunities of the AI era" for Korean semiconductor companies. NVIDIA is employing a strategy of appropriately utilizing two powerful memory partners, Samsung and SK, for supply chain stability, while also inducing fierce competition among TSMC, Intel, and Samsung in the foundry sector to optimize price and quality. Amidst these changes, Samsung Electronics is maximizing its strength as the only company capable of both memory and foundry, penetrating deeply into various AI accelerator ecosystems, including NVIDIA's next-generation platform, "Vera Rubin." This demonstrates the process of the Korean semiconductor industry positioning itself as a strategic hub that designs and produces the core infrastructure of AI technology, rather than just being a component supplier. In the end, future success depends on how quickly Samsung Electronics can secure yields for next-generation processes and prove its ability to produce customized chips that meet NVIDIA's complex requirements.