Marvell surges over 5% as Google explores co-developing AI memory and TPU chips
Google is in talks with Marvell to co-develop a memory processing unit alongside its TPU and a new TPU, with design finalization targeted next year. Marvell counts Amazon and Microsoft as custom chip clients and recently secured a $2 billion Nvidia investment while targeting $15 billion revenue in fiscal 2028.
1. Partnership Specifics
Google is negotiating with Marvell to jointly develop a memory processing unit designed to complement Google's existing tensor processing units and a new TPU for AI inference, marking a potential shift from its long-standing Broadcom TPU supplier relationship.
2. Production Timeline
The companies aim to finalize the memory chip design by next year before progressing to test production phases, with the TPU design to follow a similar schedule pending approvals.
3. Marvell's Strategic Position
Marvell already serves hyperscale customers Amazon and Microsoft with custom chips, expects to approach $15 billion in revenue by fiscal 2028, and attracted a $2 billion strategic investment from Nvidia to bolster its AI data center portfolio.
4. Market Implications
Securing a marquee partnership with Google could diversify Marvell's customer mix, challenge Nvidia's AI accelerator dominance, and position Marvell as a key player in AI hardware as demand for specialized chips grows.