Waymo Unveils In-House Chip to Power Next-Gen Robotaxis

Alphabet's Waymo has developed an in-house chip to efficiently process sensor data and run advanced AI models in real-time. The five-nanometer application-specific integrated circuit (5-nm ASIC) can deliver over 1,000 trillion operations per second of machine learning performance, improving low-light perception by cleaning up noisy camera data.

The chip, which is already in production, was built to process and combine raw information from a vehicle's cameras, radar, and lidar before feeding it into an AI system. Waymo's next-generation Ojai robotaxis have become available to the public in Los Angeles, Phoenix, and San Francisco.

Alphabet expanded its partnership with Marvell on Wednesday to develop chips for its tensor processing unit (TPU) ecosystem. This move continues to demonstrate Waymo's commitment to developing more of its own hardware, while also working with other companies such as Nvidia, AMD, Micron, Samsung, SanDisk, Socionext, and TSMC.