Google is negotiating with Samsung Electronics to produce a key part of its upcoming Tensor Processing Unit, Icefish. The chip will split manufacturing between Taiwan Semiconductor Manufacturing Company (TSMC) and Samsung’s advanced 2-nanometer technology.
Division of Production Responsibilities
TSMC will fabricate the main compute die of Icefish, while Samsung is set to manufacture the interconnect module. This interconnect links the processor to memory and plays a vital role in communication speed within the chip.
Google has developed its own TPUs since 2016 to decrease reliance on third-party GPUs. These chips support various Google services, including search and its Gemini AI models.
Samsung’s involvement marks a significant milestone for its 2nm manufacturing capabilities, which have been under development throughout 2026. Landing a contract with Google would reinforce Samsung’s position in cutting-edge foundry technology.
With Tesla already among Samsung’s custom silicon clients, Google’s extensive data center use of TPUs could translate into substantial production volumes if the collaboration advances beyond talks. Currently, discussions remain non-binding, with no official announcements made.
This partnership could shift the competitive landscape, particularly affecting Nvidia’s dominance in AI GPUs. If Google can increasingly rely on custom chips like Icefish, it may reduce its dependence on Nvidia hardware, impacting AI infrastructure suppliers.
Google’s plans for future AI chips indicate a broader push towards proprietary hardware solutions.
Investors should monitor whether Google and Samsung formalize their cooperation as it could accelerate adoption of 2nm chip technology industry-wide.
This article provides information and should not be considered financial advice.
