Google Partners with Marvell to Build New AI Chips
Google Partners with Marvell to Build New AI Chips
Google is in talks with Marvell to design two AI chips, including a memory-processing unit for TPUs, signaling a push to compete with Nvidia in AI hardware.
Alphabet's Google is reportedly in discussions with Marvell Technology to develop two new chips designed to boost how efficiently AI models run.
One chip may be a memory processing unit intended to work alongside Google's Tensor Processing Units (TPUs), while the other could be a new TPU built specifically for running AI workloads. The plan appears to aim for finalizing the MPU design by next year and then moving toward production testing, underscoring Google's ongoing push to strengthen its AI accelerator lineup.
This development fits into Google's broader effort to expand its semiconductor ecosystem, including expanding partnerships with Intel, Broadcom, and others to improve chip design and manufacturing capabilities. The goal is to ensure TPUs remain a competitive option for customers and to support the growing demands of AI-driven infrastructure, which has helped drive a meaningful portion of Google's cloud revenue.
The potential partnership could intensify competition with Nvidia, the dominant GPU maker in AI workloads, which has its own plans for new AI inference chips developed with technologies from Groq. By advancing MPU and TPU designs, Google aims to position TPUs as a viable alternative to Nvidia GPUs, potentially reshaping the AI chip landscape and challenging existing supply dynamics for cloud providers.
If these plans come to fruition, Google could offer even more specialized hardware for AI workloads, expanding its cloud offerings and giving customers a broader mix of accelerators to choose from. The move signals a continued strategic push by major tech players to own more of the AI hardware stack, from design to production.
