Google is reportedly in discussions with South Korean tech giant Samsung Electronics to manufacture a critical portion of its next-generation artificial intelligence processor. The Silicon Valley giant plans to split the production of its upcoming Tensor Processing Unit (TPU)—codenamed "Icefish"—between the world's leading semiconductor manufacturers, according to a report by The Information citing two people familiar with the matter.
Under the proposed multi-year strategy, Taiwan Semiconductor Manufacturing Co. (TSMC) will remain responsible for building the main computing core of the "Icefish" processor. Meanwhile, Samsung is being tapped to manufacture a specialized component that connects the main processor to its memory system, utilizing its highly advanced 2-nanometer (2nm) production technology. Google is currently collaborating with chip design firm MediaTek on the blueprint for "Icefish." The advanced AI chip is still in its early development phase, with mass production slated to begin as early as 2028.
A Major Victory for Samsung's 2nm Ambitions
Securing a manufacturing contract with Google would mark a significant triumph for Samsung’s foundry business, which has been aggressively expanding its contract chip-manufacturing division to compete with market leader TSMC. Samsung's cutting-edge 2nm manufacturing process allows engineers to pack significantly more transistors and computing power into smaller physical chips. This technological leap dramatically improves processing speeds, lowers energy consumption, and supercharges underlying AI capabilities.
The potential Google deal follows a string of manufacturing victories for the South Korean electronics giant. In July 2025, Samsung locked in a massive $16.5 billion contract with Tesla to manufacture custom automotive AI chips using the same 2nm architecture. To keep up with this growing client roster, Samsung announced it is actively considering building a second semiconductor factory in Texas to ramp up domestic U.S. production.
Breaking the TSMC Bottleneck
The move highlights an industry-wide push by major tech firms to reduce their overwhelming reliance on TSMC. While TSMC remains the premier manufacturer for global tech hardware, the historic surge in AI demand has pushed the Taiwanese foundry's capacity to its absolute limits, turning it into a potential supply-chain bottleneck for the entire tech sector. Google is actively diversifying its production lines to mitigate this risk. Just days prior, reports surfaced that Google was also in separate talks with Intel to manufacture more than three million internal TPUs by 2028.
Google’s custom, in-house AI chips have quickly emerged as one of the few viable alternatives to Nvidia’s highly dominant graphics processing units (GPUs). Sales of these proprietary TPUs have become a primary revenue driver for Google's cloud computing division. The company recently expanded its silicon lineup by unveiling two new custom chips engineered specifically to handle heavy AI model training and data inference.
At publication time, Samsung Electronics declined to comment on the manufacturing talks, while Alphabet did not immediately respond to requests for comment.
