Google to partner with Marvell to develop new AI chips

Alphabet's Google is reportedly in discussions with Marvell Technology to develop two new chips aimed at enhancing the efficiency of running artificial Intelligence (AI) models. According to  The Information report, two individuals familiar with the matter suggested that one of the chips may consist of a memory processing unit intended to work alongside Google's tensor processing unit (TPU), while the other could be a new TPU specifically designed for running AI models.

Google's initiative is part of its strategy to position its TPUs as a viable alternative to Nvidia's widely used GPUs. The sales of TPUs have increasingly become a significant contributor to the growth of Google's cloud revenue, as the company seeks to demonstrate to investors that its investments in AI are yielding returns. The Information's report also noted that the company aims to finalise the design of the memory processing unit by next year, before moving on to test production.

This development underscores Google's ongoing efforts to bolster its AI capabilities and expand its footprint in the competitive tech landscape. Recently, the tech giant has been expanding its partnership with companies like Intel, Broadcom and others to strengthen its semiconductor ecosystem, improve chip design and manufacturing capabilities, and support the growing demands of AI-driven infrastructure.

Now, as Google scales its AI accelerators or TPUs efforts, the move may come as an emerging threat to Nvidia's long-term dominance in the market. Nvidia has been developing its own new AI inference chips using technology from Groq, and now another major player could intensify competition, potentially reshaping the AI chip landscape and challenging Nvidia’s lead in high-performance computing.

Google's first-quarter earnings are also expected to be released on April 29, which could set the tone for its AI investment strategy, and will provide insights into its cloud and advertising performance, and signal how aggressively it plans to compete in the AI and semiconductor space.