Recently, according to a report by Korean media outlet "BusinessKorea," Samsung Electronics announced a major move during its Management Committee meeting:
the decision to invest in the field of Graphics Processing Units (GPUs).
This decision marks a significant shift in Samsung's internal agenda.
Previously, Samsung's discussion topics primarily focused on factory construction and equipment investments in areas such as memory semiconductors and foundry services.
Although the specific investment details have not yet been disclosed, the announcement has elicited strong reactions from the industry.
Many speculate that Samsung may aim to enhance its competitiveness in the GPU-related business field,
while others believe that Samsung's investment is actually intended to leverage GPUs to strengthen semiconductor process innovation rather than focusing on developing and manufacturing GPUs.
Specifically, Samsung Electronics can not only foundry produce GPUs (such as collaborating with AMD to incorporate its GPU technology into its own Exynos processors) but also develop and produce high-bandwidth memory (HBM).
With these experiences and technological accumulations, Samsung Electronics already possesses considerable strength in the GPU field.
By investing in the GPU field, Samsung can further consolidate and expand these businesses, thereby increasing its profitability.
At NVIDIA's "GTC 2024" conference in March this year, Samsung clearly stated that it would continue to cooperate with NVIDIA to develop AI-based digital twin systems for fully automated semiconductor factories by 2030.
According to Samsung Electronics' corporate governance report, the Management Committee of Samsung Electronics' Board of Directors approved the "GPU Investment Proposal" on March 19 this year.
The Management Committee, responsible for reviewing major matters of Samsung Electronics, comprises several senior executives from Samsung,
including Samsung Electronics Vice Chairman and Co-CEO Han Jong-hee, who is also head of the Consumer Electronics Division, along with executives from the Mobile Experience and Memory Business divisions.
Graphics Processing Units (GPUs), also known as display chips, are microprocessors used for graphics processing, image rendering, video decoding/encoding, and parallel computing.
With its powerful parallel computing capabilities, the GPU has become a main tool for training AI large models.
Currently, major global GPU manufacturers include NVIDIA, AMD, and Intel.
High-bandwidth memory chips (HBM) are an important component of GPUs.
At the end of 2022, with the rapid rise of generative AI represented by ChatGPT, the HBM market heated up, prompting large global tech companies to invest in this field.
SK Hynix became a major beneficiary in this field. Among the three major global memory semiconductor giants (Samsung, SK Hynix, Micron), SK Hynix was the first and only manufacturer to achieve mass production of HBM3 at that time.
Naturally, Samsung was not to be outdone.
Just last month, Samsung bet on its 12-layer HBM3E, launching a head-to-head competition with SK Hynix.
Samsung established a dedicated HBM R&D team to improve the yield and quality of HBM3E chips to meet NVIDIA's needs and thereby snatch SK Hynix's share in NVIDIA's transactions.
In fact, Samsung had already ventured into GPU development before.
Previously, the system LSI business division of Samsung Electronics collaborated with AMD to jointly develop GPU products for smartphones.
In 2019, Samsung Electronics and AMD first announced a "multi-year strategic partnership" agreement.
According to the agreement terms, Samsung obtained the license for AMD's RDNA graphics architecture.
Three years later, the two jointly developed a GPU based on AMD's RDNA 2 architecture (named Xclipse).
This was also the industry's first mobile GPU with hardware-accelerated ray tracing and variable rate shading, integrated into Samsung's Exynos 2200 processor.
In April last year, Samsung and AMD announced the renewal of their agreement to bring multiple generations of high-performance, ultra-low-power AMD Radeon graphics solutions to Samsung's Exynos processor product line.
The renewal of the partnership indicates the possibility of extending the collaboration from mobile to the automotive sector.
Returning to Samsung Electronics' investment in GPUs, although the official specifics have not been clearly stated, some recent arrangements reveal hints.
As semiconductor processes become increasingly complex, Samsung Electronics' Device Solutions division is dedicated to building specialized IT infrastructure,
including constructing large data centers to strengthen data-driven semiconductor process innovation capabilities.
Currently, Samsung Electronics is preparing to launch its newly constructed "Hwaseong High Performance Computing (HPC) Center" in the Hwaseong campus in Gyeonggi Province, South Korea.
This center, completed in April this year, is used to store and analyze data generated during semiconductor manufacturing and will play a leading role in the AI-based design and process automation transformation.
The Hwaseong High Performance Computing Center is equipped with large servers and network equipment necessary for semiconductor design and development.
Clearly, Samsung Electronics' investment and introduction of a large number of GPUs can meet the needs of semiconductor design and production, aiding in achieving innovation in semiconductor production processes, and improving production efficiency and product quality.
Moreover, Samsung Electronics plans to build digital twin chip factories, which also require large-scale GPU computing resources.
Undoubtedly, Samsung Electronics' substantial investment in GPUs indicates that AI computing is becoming increasingly important in its strategy.
Through this initiative, Samsung hopes to secure a favorable position in future technological competition, further advancing its development in the semiconductor and AI fields.
Comments