Investing.com -- Alphabet’s Google is in talks with Marvell Technology to develop two new chips designed to run AI models more efficiently, The Information reported Sunday, sending the chipmaker’s shares higher in premarket trading today.

Marvell stock jumped 6.3% by 04:38 ET (08:38 GMT).

One of the chips is a memory processing unit intended to work alongside Google’s tensor processing unit (TPU), while the other is a new TPU built specifically for AI inference, the report said.

The companies aim to finalize the memory chip’s design as soon as next year before moving to test production, per the report.

The talks are the latest sign of Google’s push to establish its TPUs as a credible alternative to Nvidia’s dominant graphics processing units. TPU sales have become an increasingly important contributor to Google Cloud’s revenue as the company works to demonstrate returns on its AI spending.

For most of its history, Google kept the bulk of its TPUs for internal use. That changed in 2022, when its cloud unit successfully lobbied to take over the group responsible for selling the chips externally, significantly increasing its allocation.

As customer demand for AI computing has grown, Google has moved to scale up both production and external sales of TPUs, and last year began selling them directly into customers’ own data centers rather than restricting access to its cloud platform.

Earlier this month, Google officially announced TorchTPU, an initiative to make its chips natively compatible with PyTorch, the world’s most widely used AI software framework.

The project aims to let developers migrate existing PyTorch workloads to TPUs with minimal code changes — removing a key friction point that has slowed adoption among the many customers who have already built their infrastructure around the software, and further pressuring Nvidia’s hold on the AI computing market.