Google is set to replace Broadcom with MediaTek as its design partner for the 7th-gen AI TPU chip. With strong ties to TSMC, MediaTek could help Google lower costs while improving AI performance. Find out how this shift impacts AI hardware and Google Cloud customers.
Google is reportedly shifting its AI chip strategy by replacing Broadcom with MediaTek as the new design partner for its Google AI chip—the seventh-generation Tensor Processing Unit (TPU). These TPUs are specialized AI accelerators designed to handle machine learning workloads for Google and its cloud customers.
While Google has relied on Broadcom for TPU development, there are compelling reasons behind this potential transition. MediaTek’s strong relationship with TSMC, the world’s largest chip foundry, could allow Google to lower production costs. Reports suggest Google spent between $6 billion and $9 billion on TPUs last year, and reducing these costs is a major priority.
Why Google Prefers MediaTek Over Broadcom
Despite this shift, Google is not severing ties with Broadcom entirely. However, the move toward MediaTek makes sense for several reasons:
- Lower Manufacturing Costs – MediaTek’s collaboration with TSMC allows for more competitive pricing compared to Broadcom.
- Reduced Dependence on Nvidia – Google’s AI chip strategy aims to minimize reliance on Nvidia’s GPUs, which are the dominant choice for AI training but can lead to shortages and high costs.
- Better Supply Chain Management – Working with MediaTek could streamline Google’s supply chain, preventing bottlenecks in AI chip production.
Google’s AI Strategy vs. Rivals
Google’s AI chip approach differs significantly from that of its competitors like OpenAI and Meta Platforms, which remain heavily reliant on Nvidia GPUs. This dependency has led to issues—such as OpenAI running out of Nvidia GPUs, forcing delays in releasing its GPT-4.5 model.
Unlike CPUs, which process tasks sequentially, GPUs excel at parallel processing, making them ideal for AI workloads. However, with custom TPUs, Google aims to optimize AI performance while reducing reliance on third-party hardware.
What This Means for AI Hardware
The potential shift to MediaTek for Google AI chip design could bring:
- More cost-efficient AI hardware for Google Cloud customers
- Greater independence from Nvidia in AI chip production
- A competitive edge in the AI race against OpenAI and Meta
As Google continues to invest in AI hardware, this move signals a strategic effort to control costs and enhance efficiency. If the partnership with MediaTek is confirmed, it could reshape the landscape of AI chip development.
