Google is negotiating with Marvell Technology to develop two custom AI chips. The collaboration focuses on a memory processing unit for Google’s Tensor Processing Units and a dedicated AI inference chip. These components aim to improve the efficiency of running large-scale AI models.
The initiative signals Google's intent to diversify its hardware supply chain. This move reduces dependence on Broadcom, despite a recently renewed multi-year agreement between the two companies.
A partnership with a major cloud provider would strengthen Marvell’s standing in the custom silicon market. Marvell currently competes to provide specialized hardware for global hyperscalers.