"Focus on TSMC-SK Relationship"
If SK Maintains Leadership in HBM Technology,
Triangular Alliance System Likely to Continue for Now
As SK Hynix achieved its highest-ever performance in the third quarter, the 'Team Nvidia' triangular alliance consisting of the U.S. Nvidia and Taiwan's TSMC is gaining attention. This is because the collaboration between Nvidia's artificial intelligence (AI) accelerators, SK Hynix's high-bandwidth memory (HBM), and TSMC's packaging is generating strong synergy effects. With big tech companies planning to continue investing in AI, analysts predict that Team Nvidia will lead the global AI semiconductor market and continue to thrive.
The synergy effects of these companies can be seen in their recent performance and stock prices. SK Hynix, which recorded its highest-ever third-quarter earnings, saw its stock price rise by 54.6% over the past year, and its global HBM market share reached 53%.
Nvidia also exceeded market expectations with an operating profit of $18.642 billion (approximately 24.9336 trillion KRW) in the second quarter (May to July). The company's stock price increased by 224.7% over the past year. Notably, on the 21st (local time), Nvidia's stock price traded on the New York Stock Exchange reached $141.80, surpassing its previous all-time high of $140.89. Its market capitalization also grew to $3.474 trillion, closely approaching Apple’s $3.588 trillion, the market cap leader at the same time.
TSMC, the world's number one foundry (semiconductor contract manufacturing) company, posted a net profit of 325.2 billion Taiwan dollars (approximately 13.9 trillion KRW) in the third quarter, a 54.2% increase compared to the same period last year. This figure exceeded the consensus estimate of 300.2 billion Taiwan dollars. Its stock price also rose by 94.9% over the past year.
The triangular cooperation of Team Nvidia is expected to strengthen further. IT big tech companies such as Microsoft, Amazon, Meta (Facebook's parent company), and Alphabet (Google's parent company) plan to increase AI-related investments next year, which is likely to boost demand for Nvidia AI accelerators.
Microsoft recently announced plans to raise over $30 billion (approximately 40 trillion KRW) with the world's largest asset management firm BlackRock to invest heavily in data centers and AI infrastructure facilities. Amazon, the world's number one cloud company, also plans to invest more than $100 billion (approximately 138 trillion KRW) in data centers over the next decade. Meta is investing $40 billion (approximately 54.46 trillion KRW) in AI this year, and Google is establishing AI data centers in places like South Carolina, USA ($3.3 billion, approximately 4.3 trillion KRW). This underlines the expectation that AI semiconductor demand will remain strong next year.
Nvidia is currently receiving SK Hynix's 5th generation HBM, the HBM3E 8-stack product. Jensen Huang, Nvidia's CEO, recently said, "Demand for Blackwell is insanely high, and every company wants to have Blackwell first." Wei Zhejia, TSMC's CEO, recently stated, "The development of AI application devices has brought the PC and mobile semiconductor markets into a recovery phase," adding, "The AI era is just beginning."
SK Hynix has also expressed its intention to further strengthen its alliance with TSMC and others. Kim Woo-hyun, SK Hynix's Chief Financial Officer (CFO) and Vice President, said during the third-quarter earnings conference call on the 24th, "The 6th generation high-bandwidth memory (HBM), HBM4, will double the number of I/O (semiconductor information gateways) and apply a new scheme to achieve low power consumption." He emphasized, "Since many technological changes are expected, including the first use of (TSMC's) logic dies, we are collaborating as one team with our foundry partners." He added, "We are targeting customer shipments in the second half of 2025 as planned."
Professor Beom Jin-wook of Sogang University's Department of Electronic Engineering and System Semiconductor Engineering said, "Since Team Nvidia consists of three top-tier AI semiconductor companies, they will widen the gap with second-tier companies through internal information exchange," adding, "From Nvidia's perspective, dealing with SK, which leads in 5th generation HBM3E 12-stack and above products and 6th generation HBM4, is a way to enhance production efficiency and synergy."
From TSMC's perspective, there is also analysis that, as it competes with Samsung Electronics' foundry business, it prefers collaboration with SK memory over Samsung memory (HBM).
Kim Hyung-jun, head of the Next-Generation Intelligent Semiconductor Business Group and honorary professor at Seoul National University, said, "From Nvidia's standpoint, they are so desperate to increase sales of the new AI accelerator product Blackwell that they have asked TSMC to build a dedicated fab for Nvidia," adding, "They would prefer to continue receiving SK HBM rather than Samsung's." He continued, "Since TSMC has maintained a close relationship with SK Hynix, sharing data for a long time, it is inevitable that they prefer SK HBM."
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
