Google and Marvell Discuss Partnership for New AI Chips: The Information Reports

• The two chips under development are a memory processing unit designed to work with Google's existing TPUs and a new TPU built specifically for running AI models
• Companies aim to finalize the memory processing unit design by next year before moving to test production
• TPU sales have become a key driver of Google's cloud revenue growth as the company seeks to demonstrate returns on its AI investments and compete with Nvidia's GPU dominance

Summary

Key Development:

Alphabet's Google is in discussions with Marvell Technology to develop two new AI chips aimed at improving efficiency in running AI models, according to The Information report citing sources familiar with the talks.

Chip Specifications:

The partnership focuses on developing:

1. A memory processing unit (MPU) designed to work alongside Google's existing tensor processing units (TPUs)
2. A new TPU specifically built for running AI models

Timeline:

The companies aim to finalize the memory processing unit design as early as next year before proceeding to test production phase.

Strategic Context:

Google has been working to establish its TPUs as a competitive alternative to Nvidia's dominant graphics processing units (GPUs) in the AI chip market. TPU sales have become a significant growth driver for Google's cloud revenue, helping demonstrate return on investment to shareholders concerned about the company's substantial AI expenditures.

Market Implications:

This partnership represents Google's continued effort to reduce dependence on Nvidia's GPU ecosystem and strengthen its position in the AI infrastructure market. The development could impact the competitive dynamics in the AI chip sector, particularly benefiting Marvell as a key hardware partner. For investors, successful TPU adoption could validate Google's AI strategy and boost cloud division performance.

Verification:

Neither Google nor Marvell immediately responded to Reuters' request for comment, and Reuters could not independently verify the report.

The collaboration underscores intensifying competition in AI chip development as tech giants seek proprietary solutions to power increasingly demanding AI workloads.