Google's next-generation TPUv8e AI chip will reportedly utilize Intel Foundry's EMIB packaging, marking a series of wins for Intel in recent days.
Intel Foundry Might Be Making Google's Next-Gen TPU With Its Advanced Packaging "EMIB" Capabilities
Intel has been in the limelight for the past couple of days as it buisnesses gain traction from the recent surge in Agentic AI, which has led to CPUs becoming the hottest commodity after GPUs and memory.
Now, reports are coming from Taiwanese outlet, Commercial Times, that Google will leverage Intel's EMIB technology for its next-generation TPU chip. Intel Foundry has already secured Tesla as a big customer for 14A, and Tesla itself will be leveraging the process technology at its TeraFab, which is set to be a critical component of its compute requirements.
The supply chain points out that AI training chips, due to their advanced manufacturing processes and highly complex packaging, have significantly increased testing intensity.
Google recently launched its brand new TPU8i (Sunfish) and TPU8t (Zebrafish) chips, designed for Inference and Training workloads, respectively. These chips are manufactured using the TSMC CoWoS (Chip-on-Wafer-on-Substrate) technology, which has been the go-to solution for the bulk of chips powering AI facilities around the globe.
TSMC will reportedly increase its wafer allocation of TPUv8x chips from 70,000 wafers to 160,000 wafers as the initial redesign issues have been resolved, leading to risk production commencing in May this year. The new wafer allocation will lead to 3.2 million CoWoS packaged chips.
For its next-generation Google TPUv8e, Intel is said to be the one who will be producing the chip using its EMIB (Embedded Multi-Interconnect Bridge) technology. Compared to CoWoS (2.5D), EMIB is said to be far more versatile, offering low-cost and scalable designs. TPUv8e is expected to have its main compute die made by Google itself, while the I/O and back-end design will be made by MediaTek.
For TPUv8p, Google is expected to once again rely entirely on Broadcom for the compute, IO, and back-end design, while the chip will be made using TMSC's CoWoS/SoCI packaging technologies. Though the use of Intel Foundry is not official, Intel itself has stated that it won't name any customer itself & will rather wait for an announcement by its customers and their respective products.
The next-gen TPUv8 chips codenamed Humufish are expected to reach 3.5 million units by the end of 2028.
With an expected timeline of Q4 2027 for both Google TPUv8e and TPUv8p, we should be hearing an announcement around the end of this year or early next year. Google partnering with Intel will once again be a major driver of confidence for Intel Foundry as its 14A process technology and EMIB solutions are seen as the next big thing, one that might alleviate the current supply chain crisis, which has been the result of the major AI boom.
News Source: Dan Nystedt
About the author: A Software Engineer by training and a PC enthusiast by passion, Hassan Mujtaba serves as Wccftech's Senior Editor for hardware section. With years of experience in the industry, he specializes in deep-dive technical analysis of next-generation CPU and GPU architectures, motherboards, and cooling solutions. His work involves not only breaking news on upcoming technologies but also extensive hands-on reviews and benchmarking.
Follow Wccftech on Google to get more of our news coverage in your feeds.
