Waymo lifts the lid on the ‘brain’ powering its robotaxis

Autonomous driving developer Waymo has publicly shared comprehensive technical specifications of the high-performance computing hardware installed in its driverless fleet.

Located in the trunk of each vehicle, this powerful system integrates custom-built silicon and specialized processors designed to handle massive flows of real-time data from various onboard sensors, including cameras, radar, and lidar. The core architecture prioritizes rapid response times, durability against environmental stressors, and physical safety backups. To process this complex data stream efficiently, the company introduced a specialized five-nanometer application-specific integrated circuit that pre-processes and merges raw inputs before they reach the main computer, significantly reducing processing latency.

While Waymo manages the core system design and machine learning framework internally, it relies on global semiconductor and technology suppliers like Nvidia, TSMC, and Samsung for individual components. This collaborative approach highlights a sharp divergence from competitors who advocate for camera-only systems, as Waymo asserts that sensor redundancy is essential for safe, scaled driverless operations. However, despite successfully shrinking both the size and the production cost of these trunk-mounted units in its latest hardware generation, the overall expense of outfitting each vehicle with such sophisticated computing power remains a significant financial hurdle as the company attempts to transition from a technological milestone to a profitable business model.