Waymo, Alphabet's autonomous vehicle unit, has unveiled insights about the custom semiconductor technology driving its fleet of self-driving taxis across multiple U.S. cities, highlighting the balance of performance, reliability, and automotive-grade standards.
- Dual custom ASICs execute over 1,000 trillion calculations per second.
- TSMC 5nm automotive node ensures production quality and reliability.
- Modular system processes data from multiple sensors including cameras, lidars, and radars.
What happened
Waymo has provided detailed information about the sixth version of its autonomous driving compute module, which incorporates two custom-built application-specific integrated circuits (ASICs). These chips deliver immense computing power, capable of exceeding 1,000 trillion calculations per second, to manage the demands of real-time vehicle operation.
The chips are manufactured using Taiwan Semiconductor Manufacturing Company’s five-nanometer automotive-grade process, known as N5A, which is tailored for enhanced reliability and compliance with automotive industry standards. Waymo has configured these ASICs to run AI models specialized in sensor fusion—combining data from 13 cameras, four lidars, and six radars to create a precise environment map for safe navigation.
Why it matters
The use of highly specialized ASICs optimized for low-latency onboard AI inference represents a significant step in autonomous vehicle technology. By processing sensor data entirely within the vehicle, Waymo avoids latency issues associated with cloud-dependent computation, critical for safety in fast-moving traffic situations.
Including two redundant processors in the module adds fault tolerance, enabling the system to maintain operation even if one processor fails. This level of hardware resilience, combined with heightened environmental durability through liquid cooling and robust enclosure design, underpins the system’s ability to operate reliably in challenging conditions including vibration, shock, and temperature extremes.
What to watch next
Observers will be interested to see how this sixth-generation compute module impacts Waymo's operational efficiency and safety performance across its current fleet of around 4,000 vehicles in 11 U.S. cities. The integration with multiple vehicle models, such as Jaguar I-Pace crossovers and Hyundai Ioniq 5 SUVs, also highlights Waymo's platform adaptability.
Further developments to watch include advances in ASIC design and manufacturing that might allow even greater performance, efficiency, or new AI capabilities for autonomous driving. Industry adoption of similar specialized chips by other autonomous vehicle players will also indicate how crucial custom silicon continues to be in competing for leadership in this emerging market.