Google is in discussions with Marvell Technology, the semiconductor company, to jointly develop two new chips intended to improve the efficiency of running artificial intelligence models, The Information reported, citing two people with knowledge of the talks.
One chip is described as a memory processing unit designed to work alongside Google's existing tensor processing units (TPUs), targeting the data movement bottleneck that constrains performance in large-scale AI workloads.
The other is a new TPU built specifically for inference, the phase of AI computing in which trained models serve users rather than learn from data.
The companies aim to finalise the design of the memory processing unit as soon as next year before handing it off for test production.
The discussions have not yet produced a signed contract, and neither Google nor Marvell responded to requests for comment.
The talks would add Marvell as a third design partner in Google's custom silicon supply chain alongside Broadcom, which recently locked in a TPU agreement running through 2031, and MediaTek, which is reportedly working with Google on a next-generation data centre chip.
Google has been pushing to establish its TPUs as a viable alternative to Nvidia's dominant GPUs, and sales of the in-house chips have become a key driver of growth in its cloud division.
The company secured a significant endorsement in February when Meta signed a multi-year deal to lease Google's TPU-powered infrastructure for training advanced AI models.
Related reading
- Nvidia launches open-source quantum AI models as it deepens grip on chip design workflows
- Meta partners with Broadcom on MTIA AI chips
- Nvidia says the open versus proprietary AI debate is the wrong argument
Marvell's data centre revenue reached a record $6.1 billion in its fiscal year ending February 2026, with total revenue of $8.2 billion, up 42% year on year.
The company runs a custom silicon business with an annual run rate of $1.5 billion across 18 cloud provider design wins, building chips for Amazon, Microsoft and Meta in addition to its existing work with Google on the Axion ARM-based processor.
The recap
- Talks involve Google partnering with Marvell to develop chips
- One chip described as a "memory processing unit" for TPUs
- Design expected to be finalised as soon as next year
