The autonomous driving unit developed a specialized 5 nm processor to handle raw data from 13 cameras, lidar and radar sensors in real time. This ASIC delivers 1,000 TOPS of machine learning performance, reducing latency from the first pixel to action to facilitate onboard driving decisions within milliseconds. The system employs a redundant architecture designed as two independent engines to ensure safety without a human backup.
Vertical integration of the 6th generation system lowered the estimated hardware cost per vehicle to a range of $20,000 to $25,000, down from $100,000 to $125,000 in the previous version. Waymo uses a heterogeneous compute stack that incorporates parts from Nvidia, AMD, Micron, Samsung and TSMC. The total processing power of the onboard system has increased 20 times over the past eight years.
Key sources
- SOURCE@sawyermerritt“The ASIC’s specialized accelerators instantly extract critical information from raw lidar, radar, and camera streams”x.com
- SUPPORT@wallstengine“can process all 13 high-resolution cameras simultaneously in real time”x.com
- SUPPORT@wallstengine“down from ~$100K-$125K for the previous generation”x.com
- SUPPORT@stocksavvyshay“A big driver is vertical integration including a custom 5nm chip capable of 1,000 TOPS”x.com
- SUPPORT@edludlow“Waymo's purpose-built 5nm ASIC”x.com
- SOURCE@stocksavvyshay“$GOOGL Waymo has cut estimated robotaxi hardware costs from ~$115K per vehicle to just ~$20K with its 6th-gen system”x.com
- SUPPORT@roelsmelt“That changes the operating model: more cities, more vehicles, faster learning.”x.com
- SUPPORT@trungtphan“Original Jaguar I-PACE fleet costs $200k per vehicle”x.com