After falling behind SK Hynix, Samsung begins leveraging its financial prowess.
As the largest memory chip manufacturer in the industry, Samsung Electronics faces significant pressure in the HBM market.
Previously, due to issues such as excessive power consumption, overheating, and low yield rates, Samsung’s latest HBM3E memory failed NVIDIA’s tests, putting them at a disadvantage against SK Hynix and Micron.
Additionally, due to internal management issues, Samsung made a rare move by urgently reshuffling its semiconductor department mid-year, hoping to adjust the internal atmosphere within the department.
However, according to the latest news, Samsung’s HBM still has a chance to supply NVIDIA, and they have refuted previous news reports.
Interestingly, according to Korean media outlet BLOTER, Samsung Electronics’ Board Management Committee approved a proposal to invest in GPUs earlier this year.
Although the specific content and direction have not been disclosed, it is evident that the top management of Samsung Electronics has reacted quite quickly to HBM chip developments.
01
First-time disclosure of GPU investment in 12 years
According to the latest disclosed Samsung Electronics corporate governance report, the Board Management Committee approved a “GPU investment proposal” in March this year, with top management deciding to invest in graphics processing unit (GPU) related technologies.
The mentioned management committee, led by Han Jong-hee, head of Samsung Electronics’ other major business division, Consumer Electronics (DX), includes key executives such as Roh Tae-moon, president of the Mobile Business (MX) division, Park Hak-kyu, head of the Business Support Office, and Lee Jung-bae, president of the Memory Business. The recently replaced head of the Device Solutions (DS) department, Kyung Kye-hyun, also attended the meeting that decided on the GPU investment.
Typically, the agendas disclosed by the management committee focus mainly on factory construction and equipment investment in the storage and logic fields.
But this time is unusual. Since the public disclosure of agendas began in 2012, this is the first time a GPU-related business field topic has appeared.
In fact, the news of Samsung investing in GPUs is not entirely unexpected. In recent years, with the surge in demand for AI semiconductors, chip manufacturers have been making investment decisions related to advanced packaging, and chip giant Samsung is no exception.
However, due to the ambiguous language in the corporate governance report, it is currently difficult to determine the specific direction of Samsung Electronics’ “GPU investment” decision, so BLOTER provided two general directions.
First, Samsung is likely using this investment to directly enhance its competitiveness in GPU-related businesses, including development and manufacturing.
Currently, general-purpose GPUs are mainly used for AI computing, with their configured HBM memory provided by memory manufacturers like Samsung and SK Hynix. Additionally, Samsung’s System LSI division is working with AMD to develop GPUs for smartphone application processors (AP), which is also a promising business area for the foundry division.
Considering Samsung’s investments in the HBM field and foundry over the past few months, it is clear that top management is increasingly focusing on AI.
However, some analysts believe that the investment is not aimed at helping Samsung develop and manufacture GPUs but rather at deepening process innovation to purchase more GPUs.
In March this year, Samsung announced at “GTC 2024” that it would collaborate with NVIDIA on AI-based initiatives to build a fully automated semiconductor factory by 2030.
Currently, Samsung Electronics’ new High-Performance Computing (HPC) center in Hwaseong, Gyeonggi-do, was completed in April this year. At the same time, the DS department has built dedicated IT infrastructure at its manufacturing plants both domestically and overseas.
However, as semiconductor engineering becomes increasingly detailed and complex, the role of AI computing in enhancing data utilization becomes more prominent.
At this point, purchasing a large number of GPUs necessary for AI computing will incur significant costs.
02
Conclusion
From Samsung’s recent actions, both routes proposed by the Korean media are possible.
On one hand, Samsung’s foundry is taking active measures to win NVIDIA’s next-generation GPU 3nm orders. But constrained by yield rate issues, the vast majority of AI chips are currently contracted to TSMC alone. To take on NVIDIA’s business, upgrading its own foundry lines is essential.
🔼 Samsung and TSMC (Image Source: Internet)
On the other hand, Samsung lacks the capacity to produce products that can replace NVIDIA GPUs in the short term.
Therefore, investing funds to purchase NVIDIA GPUs is almost the only option at present.
It can be said that in the face of the AI wave, no one can escape GPU anxiety, not even a powerhouse like Samsung Electronics.
Disclaimer: This article is created by the original author. The content of the article represents their personal opinions. Our reposting is for sharing and discussion purposes only and does not imply our endorsement or agreement. If you have any objections, please get in touch with us through the provided channels.
