Waymo Releases New Custom Chip

Waymo has introduced its 5nm ASIC, a purpose-built chip engineered exclusively for the real-time conversion of raw sensor data from robotaxis’ lidar, radar and camera sensors before they reach the vehicle’s core machine learning brain. The chip can deliver more than 1,000 trillion operations per second, ensuring that driving commands can be executed by the vehicle in milliseconds. The chip is just one step in Waymo’s move away from off-the-shelf components towards custom systems to operate its vehicles.

The 5nm ASIC contributes to the responsiveness of Waymo’s vehicles, one of its stated “three non-negotiable requirements”. To reach the level of responsiveness necessary for the operation of autonomous vehicles requires massive amounts of raw compute power. The chip represents a 20x scale-up in such raw compute power by Waymo over the past eight years. In addition to improving responsiveness, Waymo continues to hone the ruggedness and redundancy of the vehicles’ machine learning brain to ensure that they perform at the highest level.

Alongside its efforts towards custom systems to operate its vehicles, Waymo continues to partner with leading technology companies that provide the powerful foundation necessary for Waymo to efficiently scale its autonomous computing system, including AMD, Micron, NVIDIA, Samsung, Sandisk, Socionext and TSMC.

The new 5nm ASIC will be featured in Waymo’s newest vehicle model, the Ojai robotaxi, which is currently in service in three US cities: Phoenix, Los Angeles and San Francisco. Waymo is now active in 11 cities across the US and has plans to add 19 more to its list. Earlier this year, Waymo announced that it had reached 500,000 paid rides per week.

Author: Samantha Curry, 2026 Articling Student