Google Develops AI Chips with Marvell

Google is reportedly in talks with Marvell Technology to develop two specialized chips for artificial intelligence (AI). One of these chips is a memory processing unit designed to work alongside Google's existing tensor processing unit (TPU). The other chip is said to be a new TPU specifically created to run AI models. Google has long used its own TPU chips to power AI systems. These TPU chips are designed especially for machine learning and neural network tasks, offering higher efficiency compared to standard GPUs.

This new chip development is seen as a move by Google to strengthen its hardware ecosystem as an alternative to NVIDIA GPUs. With the demand for chips in AI data centers rapidly increasing, companies like Google are focusing more on designing their own chips. Additionally, Google is currently collaborating with Broadcom on TPU development, and these new talks with Marvell indicate an effort to diversify chip suppliers. This approach also provides opportunities to enhance supply chain flexibility and control costs.

Meanwhile, the design of the memory processing unit is planned to be completed within the next year, with hopes to move into test production afterward. Some reports suggest that Google intends to deploy millions of such chips. Considering all these factors, competition in the AI hardware sector is intensifying, and it is clear that Google is making long-term investments to strengthen its AI and cloud services.