In response to the escalating demand for Tensor Processing Units (TPUs), Google is set to unveil a new generation of TPUs, internally dubbed 'Icefish'. The computational engine of this chip will be crafted by TSMC utilizing a cutting-edge 1.4-nanometer process. Meanwhile, the memory input/output die will be entrusted to Samsung for production. Mass production is slated to commence as early as 2028, and the chip is presently in the design phase, with Google partnering with MediaTek on this endeavor. This collaboration offers Samsung a prime opportunity to demonstrate its manufacturing prowess, as its 2-nanometer process will undergo rigorous real-world testing.
In recent years, Samsung has been proactively expanding its foundry business, securing orders from esteemed clients such as Google, Tesla, and NVIDIA. Currently, the scarcity of chip production capacity has emerged as a pressing issue. With the presence of two major memory chip giants, South Korea's influence in the AI industry continues to surge. Furthermore, NVIDIA recently inked a memory chip technology cooperation agreement with SK Hynix, further solidifying the region's technological standing.
