Via 247wallst.com
The deal covers AI, storage, and networking chips and could reshape Google's custom silicon supply chain beyond its longtime partner Broadcom.
Marvell Technology has announced discussions with Google to develop chips spanning AI, storage, and networking. The talks were first made public on April 19, 2026, and position Marvell as a design partner in Google’s custom silicon operation.
What Marvell is building for Google #
The collaboration involves two distinct custom AI chips. The first is a Memory Processing Unit, or MPU, designed to complement Google’s existing Tensor Processing Units. The second is a new inference-optimized TPU, built specifically for AI inference workloads.
Marvell’s role is as a design-services provider, supplementing the existing TPU lineup rather than replacing it. MediaTek previously worked with Google in a similar capacity on its Ironwood TPU.
The target timeline has the MPU design finalized by 2027, with test production to follow.
The Broadcom factor #
Google’s partnership with Broadcom on TPU and networking components was recently extended through 2031.
When reports of the Google-Marvell talks first surfaced on April 19, 2026, Marvell’s stock jumped nearly 5-6%. Broadcom’s shares, meanwhile, dipped slightly.
Marvell’s broader AI play #
Nvidia made a $2 billion equity investment in Marvell, focused on AI accelerators and interconnect technology.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our