Waymo unveils custom 5nm AI chip for robotaxi sensor fusion

Waymo, an Alphabet subsidiary, has unveiled its first custom application-specific integrated circuit (ASIC) designed to power its next-generation robotaxi platform, Ojai. Manufactured using TSMC’s 5-nanometer process, the chip is purpose-built for sensor fusion, processing real-time data from 13 cameras, four lidars, and six radars to minimize latency in driving decisions.

The custom silicon delivers over 1,000 trillion operations per second (TOPS) of machine learning performance. It is specifically engineered to handle edge ingestion tasks, such as temporal denoising to improve visibility in low-light environments. To ensure safety and reliability, Waymo utilizes a redundant architecture where two ASICs operate in parallel; if one fails, the other can immediately take over.

While Waymo is moving toward in-house silicon for latency-critical sensor processing, it continues to rely on third-party partners for other computing needs. The company maintains partnerships with Nvidia, AMD, Micron, Samsung, SanDisk, and Socionext to provide CPUs, GPUs, and other essential components for its autonomous driving stack.