Samsung Electronics has outlined its intent to penetrate the graphics processing unit (GPU) sector in its FY 2023 Corporate Governance Report. The initiative forms part of a larger strategy to diversify its semiconductor segment and lessen dependency on third-party suppliers, sources suggest.
Considering Foundry Enhancement
One approach Samsung may adopt is boosting its foundry operations to facilitate GPU manufacturing, according to The Register. As the second-largest foundry by revenue, Samsung could benefit from the rising demand for AI accelerators. A foundry is a factory where companies like Samsung manufacture silicon wafers that are subsequently processed into integrated circuits or chips
As major cloud service providers like AWS, Google, Meta, and Microsoft are expanding their AI infrastructure, demand for AI processors is putting pressure on production capacities, particularly at TSMC, which is struggling to keep up with the demand for its CoWoS technology.
Nvidia CEO Jensen Huang has previously lauded TSMC’s capabilities, emphasizing the extensive supply infrastructures required for GPU production. By fortifying its foundry services, Samsung could position itself as an alternative to TSMC, especially given the geopolitical concerns involving Taiwan.
This goes so far as some TSMC partners, such as Dutch ASML, reportedly possess the capability to remotely disable its advanced machinery installed at TSMC facilities. The company has recently explored options to shift its production facilities from Taiwan but concluded that such a move is impractical due to logistical challenges.
Samsung In-House GPU Development
Samsung is also contemplating the creation of its own GPUs targeting mobile devices, data centers, and AI applications. The company’s Exynos System on Chips (SoCs) already incorporate GPUs, with the latest Exynos 2400 featuring an Xclipse 940 GPU rooted in AMD’s RDNA 3 architecture.
Korean outlets reports that Samsung plans to launch an Exynos chip in 2025 outperforming Qualcomm’s counterparts, with a Samsung-designed GPU potentially debuting in 2026’s Exynos lineup. This move could allow Samsung to tailor its devices to specific requirements, improving AI and gaming capabilities.
Venturing into the GPU market would mean direct competition with entities like NVIDIA and AMD. Samsung, despite its semiconductor prowess and financial muscle, faces hurdles in developing discrete GPUs. It has historically shown no inclination to become a CPU or SoC vendor, let alone enter the GPU sphere.
Samsung could potentially develop GPUs compatible with Nvidia’s CUDA software, but Nvidia has previously resisted similar moves from Chinese firms and might pose legal obstacles. Furthermore, the intricacies involved in high-end GPU development are notable, as evidenced by Intel’s challenges despite its hardware and software experience.
Government Influence and Market Dynamics
South Korea’s government has been pushing local chipmakers to diversify beyond memory manufacturing, possibly influencing Samsung’s potential GPU market entry. Growing demands for high-performance computing, driven by advances in AI, machine learning, and data analytics, offer lucrative prospects for firms capable of addressing these needs. As these technologies progress, the demand for robust and efficient GPUs is expected to rise.
While Samsung’s specific strategy remains uncertain, its dedication to expanding its semiconductor portfolio indicates a significant move into the GPU market. Details regarding Samsung’s GPU plans, including development timelines and potential partnerships, are still unknown. However, industry experts believe Samsung’s entry could intensify rivalry with existing major players, leveraging its semiconductor expertise and financial strength to produce competitive GPU offerings.
Last Updated on November 7, 2024 3:48 pm CET
