Beyond Chips: Samsung’s GPU Project Sparks Speculation

According to Business Korea, Samsung’s recently released corporate governance report reveals that its Board Management Committee approved a proposal on March 19th of this year to invest in a GPU project. While specific details remain unclear, the topics discussed differ from the typical issues in the semiconductor industry.

This marks the first GPU investment decision since the agenda item was publicly disclosed in 2012, sparking speculation about Samsung’s intention to bolster its competitiveness in the GPU-related business sector. Currently, Samsung collaborates with AMD to develop GPUs for smartphone SoCs. Additionally, Samsung is working on a GPU project for memory types like HBM to better address artificial intelligence (AI) computing needs. Previously, Samsung introduced HBM-PIM (processing-in-memory) chips, which integrate an AI processor into each memory module, offering up to 1.2 TFLOPS of embedded computing power, enabling the memory chip itself to perform operations typically handled by CPUs, GPUs, ASICs, or FPGAs.

However, some analysts suggest that this is part of Samsung’s internal strategy, aimed not at developing and manufacturing conventional GPUs, but at leveraging GPU technology to enhance semiconductor processes. At GTC 2024 in March, Samsung announced its continued collaboration with NVIDIA to develop AI-based digital twin technology for fully automated semiconductor factories by 2030, aiming to improve semiconductor manufacturing efficiency. Furthermore, Samsung is constructing a high-performance computing (HPC) center at its Hwaseong campus, which may incorporate Samsung’s proprietary GPU chips in the future.

Samsung’s substantial investment in the GPU project indicates that artificial intelligence and computing are becoming increasingly pivotal in its strategic planning.