Google in Talks with Samsung to Manufacture AI Chips, Reducing Reliance on TSMC

The Information, a US media outlet, reported on the 11th, citing two sources, that Google's parent company Alphabet is in talks with Samsung Electronics to outsource the production of some of its next-generation AI processors to Samsung. According to the report, Google plans to have TSMC manufacture the main computing core of its Tensor Processing Unit (TPU), codenamed Icefish, while Samsung would use its 2nm process technology to produce key components that connect the chip to memory. The Icefish is still in development and could enter mass production as early as 2028. Securing this order would be a major breakthrough for Samsung's foundry business. The 2nm process allows for more computing power to be integrated into a smaller area, improving chip speed, energy efficiency, and AI performance. Samsung declined to comment, and Alphabet did not immediately respond to inquiries. Reuters could not independently verify the report. In April, Samsung said it expects chips produced using advanced process technology to attract more customers and is evaluating building a second plant in Texas, USA, to expand capacity. In July 2025, Samsung secured a $16.5 billion contract to produce 2nm AI chips for Tesla. This move also indicates that Google is seeking to reduce its reliance on TSMC. As AI demand grows rapidly, TSMC continues to expand production, but the market is concerned that its capacity could become a bottleneck for the entire AI industry supply chain. The Information recently reported that Google is also in talks with Intel to outsource the production of over 3 million TPU chips by 2028. Google's in-house AI chips have emerged as an important alternative to Nvidia's GPUs in recent years, and with growing TPU sales, the business has become a key growth driver for Google Cloud.