Google is making a strategic shift in its AI hardware approach by partnering with MediaTek for its seventh generation Tensor Processing Unit (TPU). Previously, Google relied on Broadcom for designing its AI accelerator chips, which power cloud computing and data centers. This move is expected to reduce costs and enhance efficiency, leveraging MediaTek’s strong ties with TSMC for more cost-effective production.

According to reports, Google spent around $9 billion on TPUs last year, so even a slight reduction in costs per chip could result in billions in savings. While the company is transitioning to MediaTek, it will still maintain some reliance on Broadcom for certain designs. By diversifying its chip design partnerships, Google aims to reduce dependency on third party solutions like NVIDIA’s GPUs, giving it greater control over AI hardware development.

This shift is part of Google’s broader effort to strengthen its AI ecosystem and avoid supply chain constraints. More details on this collaboration are expected in the coming months.
