Broadcom is doubling down on custom AI hardware, and its latest announcement makes one thing clear: the company intends to stay deeply embedded in the next wave of data center acceleration for years to come.
In a major partnership update, Broadcom confirmed it will continue working with Google on Tensor Processing Unit (TPU) application-specific integrated circuits (ASICs) through 2031. That long runway matters because TPU chips are central to Google’s AI and cloud infrastructure, powering large-scale machine learning workloads across training and inference. By extending this collaboration across multiple future TPU generations, Broadcom positions itself to keep earning revenue tied to Google’s ongoing TPU roadmap, rather than relying on a single product cycle.
For Broadcom, long-term custom silicon agreements like this are strategically valuable. They help lock in predictable demand, strengthen relationships with hyperscale customers, and keep Broadcom at the center of the AI hardware supply chain as competition heats up in custom accelerators and data center chips.
At the same time, the competitive landscape is getting more crowded. Rival chipmakers are also chasing the same high-stakes AI infrastructure deals, with companies like MediaTek ramping up efforts to win business in advanced silicon programs. That rising pressure is exactly why multi-year agreements with major cloud players are so important: they can create stability and scale, even as the market shifts and new contenders push in.
The bottom line is that Broadcom’s extended TPU partnership with Google signals staying power in AI-focused ASIC development. With the collaboration now set through 2031, Broadcom is aiming for sustained relevance and recurring revenue as Google’s TPU ecosystem evolves across several generations of AI computing.
