Google Taps Samsung for Next-Gen AI Chip Production

Google is reportedly looking to diversify its supply chain for its proprietary artificial intelligence (AI) chips, with sources indicating a potential collaboration with Samsung Electronics for the production of its next-generation Tensor Processing Unit (TPU). This upcoming 10th-generation TPU, internally codenamed "Icefish," is a critical component for Google's expanding AI capabilities, powering services like its Gemini AI model. The move comes as the global demand for advanced AI semiconductors continues to outstrip manufacturing capacity, leading to significant pressure on leading foundries like Taiwan Semiconductor Manufacturing Co. (TSMC). Historically, Google has primarily relied on TSMC for the production of its TPUs. However, the intense competition for TSMC's advanced process technologies from major AI players like NVIDIA has prompted Google to explore alternative manufacturing partners. This potential partnership with Samsung is seen as a strategic effort to mitigate supply risks and ensure a stable supply of high-performance AI hardware for its internal needs and growing external corporate customers. Under the proposed arrangement, TSMC is expected to continue manufacturing the core computing chip of the "Icefish" TPU, utilizing its cutting-edge 1.4-nanometer process. Meanwhile, Samsung Electronics would be tasked with producing the memory input/output die (I/O Die). This I/O Die is a crucial component responsible for connecting the core processor with high-bandwidth memory (HBM), which is essential for efficiently handling the massive data workloads inherent in AI applications. Samsung's expertise in memory technologies and its 2-nanometer process capabilities are cited as key factors in Google's consideration. Mass production of the "Icefish" TPU is anticipated to commence as early as 2028. For Samsung, securing a role in such a high-profile AI chip project would be a significant win, demonstrating its advanced foundry capabilities, particularly in the competitive 2-nanometer process segment. This collaboration highlights the ongoing shifts in the semiconductor industry as tech giants seek to secure their AI hardware pipelines amidst unprecedented demand and supply chain complexities.