Taiwanese fabless semiconductor firm MediaTek has made a major claim at the Goldman Sachs Taiwan Day by outlining that for its next-generation program, it plans to only use Intel's EMIB-T packaging technology. The Taiwanese supply chain has been discussing EMIB-T regularly for months now, and MediaTek is Google's partner for the next-generation TPU custom AI chips. Intel has positioned EMIB against TSMC's CoWoS packaging technology, and MediaTek's announcement marks a strategic win for the American chip company, whose CEO has aggressively targeted the AI industry as part of his turnaround efforts.
Intel Might Have Won MediaTek Over From TSMC's CoWoS
The announcement by MediaTek comes on the heels of another report from Taiwan which claims that the firm is adding new suppliers in order to deal with anticipatory orders for chips assembled via the EMIB-T packaging technology.
Unlike CoWoS, which uses a large silicon interposer to connect different components, EMIB selectively uses silicon bridges to connect the different components of an AI package. These components include the GPU core and memory, and by removing the interposer, EMIB claims to reduce manufacturing complexity and costs.
Google's next-generation TPU custom AI chips are also reportedly considering relying on EMIB-T for their packaging, according to reports from the Taiwanese supply chain.
Next-Gen Chip Tape-out With EMIB-T Scheduled For Q4 2026, Says MEdiaTek
Now, while speaking at the Goldman Sachs Taiwan conference, MediaTek has claimed that “the next-generation program is adopting only EMIB-T, with tape-out targeted for 4Q26 and mass production by 4Q27.” While it's unclear which chips this program will cover, it is likely to involve custom AI chips and CPUs. While EMIB-T packaging is cost-efficient, the interposer used by CoWoS chips also delivers high bandwidth, which is suitable for high-end AI GPUs such as those designed by NVIDIA and AMD.
While the supply chain continues to be abuzz with reports of EMIB-T's growing popularity, well-known analyst Ming-Chi Kuo believes its eventual adaptability will depend on yield. In a detailed note, Kuo outlined that Intel had set a 98% benchmark for the process to make its progress comparable to Flip Chip Ball Grid Array (FCBGA) production. According to Kuo, Google will focus on EMIB-T yields since it was competing with NVIDIA on a cost basis. A higher yield enables more usable chips per batch, which reduces the cost incurred to produce an individual chip.
About the author: Ramish is a seasoned technology writer and editor with more than a decade of experience. He specializes in semiconductor fabrication and market analysis. With a background in finance and supply chain management - via his bachelors in Finance and a micromasters in supply chain management from MIT - Ramish combines financial rigor with deep industry insight to deliver accurate and authoritative coverage.
Follow Wccftech on Google to get more of our news coverage in your feeds.
