Alphabet's Google is in talks with Samsung Electronics to manufacture a component for its next-generation artificial intelligence processor, according to a report by The Information. The chip, codenamed Icefish and planned as Google's 10th-generation Tensor Processing Unit (TPU), would use a split manufacturing strategy with Taiwan Semiconductor Manufacturing Co. (TSMC) producing the primary compute component and Samsung handling a separate memory-interface component using its 2-nanometer process technology. The reported partnership stems from tight manufacturing capacity across the semiconductor industry, particularly at TSMC, which faces surging demand for AI chip production from companies like Nvidia, while Google's AI chips are attracting more external customers and require additional production capacity. At the time of writing, shares of Alphabet were down 1.9% in Thursday afternoon trading. Samsung has been seeking to expand its contract chipmaking business since launching its foundry operations in 2005, and securing work on Google's TPU would mark a significant validation of its advanced manufacturing capabilities in the AI chip segment.
Google Plans Split Manufacturing Strategy With TSMC and Samsung
Google has traditionally relied on TSMC to manufacture its TPUs, but the Taiwanese foundry has faced surging demand for AI chip production, particularly from Nvidia. Under the reported plan, TSMC would manufacture the most advanced portion of the chip using its upcoming 1.4-nanometer process technology, while Samsung would produce a separate component using its 2-nanometer process. Process nodes such as 2nm and 1.4nm refer to generations of chip manufacturing technology. Smaller nodes generally enable higher transistor density, which can improve performance and power efficiency. The arrangement would give Samsung an opportunity to showcase its manufacturing capabilities in a closely watched segment of the semiconductor industry.
Icefish Project Targets 2028 Mass Production With MediaTek Collaboration
The Icefish project could enter mass production as early as 2028, although the timeline remains subject to change as the chip is still in development. Google is working with Taiwanese chip designer MediaTek on the project. The chip is currently planned as Google's 10th-generation TPU for AI workloads in cloud data centers.
Samsung Foundry Secures Tesla and Nvidia Contracts Alongside Google Talks
Samsung launched its foundry business in 2005 and established a dedicated foundry division in 2017, but it has struggled to close the gap with TSMC in advanced chip manufacturing. The company won a contract from Tesla last year to manufacture its next-generation AI6 chip and is producing Nvidia's language processing unit for the upcoming Vera Rubin platform. Securing work on Google's next-generation TPU would mark a significant win for Samsung's foundry efforts.
FAQ
What is Google's Icefish chip project?
Icefish is the codename for Google's 10th-generation Tensor Processing Unit (TPU), a custom AI chip designed for cloud data center workloads. The chip is currently in development with plans for a split manufacturing approach involving both TSMC and Samsung.
Why is Google considering Samsung for chip manufacturing?
Google is exploring Samsung as a manufacturing partner due to tight capacity at TSMC, which faces surging demand for AI chip production from companies like Nvidia. Google's AI chips are also attracting more external customers, increasing the need for additional production capacity.
When could the Icefish chip enter mass production?
The Icefish project could enter mass production as early as 2028, although the timeline remains subject to change as the chip is still in development.
