Tuesday, August 25, 2026

Independent technology reporting and practical analysis

Manila ·
DEVICES

Independent reporting, useful context, and practical analysis.

Back to Technomalist
Devices / news

Waymo Built a 5nm AI Chip to Process Robotaxi Sensor Data

The purpose-built ASIC handles camera, lidar and radar data before it reaches the Waymo Driver’s central inference system.

Featured image for Waymo Built a 5nm AI Chip to Process Robotaxi Sensor Data
Featured image for Waymo Built a 5nm AI Chip to Process Robotaxi Sensor Data

Waymo has revealed a custom 5-nanometer chip designed to process the flood of sensor data generated by its autonomous vehicles.

The application-specific integrated circuit, or ASIC, sits near the front of the Waymo Driver's computing pipeline. It extracts information from cameras, lidar and radar before sending the results to the vehicle's central machine-learning system.

More than 1,000 TOPS for front-end processing

Waymo says the ASICs deliver more than 1,000 trillion operations per second of machine-learning performance. The company designed specialized accelerators for tasks including sensor fusion and temporal denoising, which can improve perception in low light.

Its latest vehicle platform processes data from 13 high-resolution cameras in real time. Waymo says co-designing the chip, sensors and algorithms helps it reduce latency and use available energy more efficiently.

Those performance and safety claims come from Waymo and have not been independently tested by Technomalist. TOPS figures also do not provide a complete comparison between autonomous-driving systems because software efficiency, workload type and achieved performance matter alongside peak hardware capability.

A redundant computer built for the road

Waymo describes its onboard compute as two independent engines running parallel workloads. If one side fails, the other is intended to take over. The hardware is liquid-cooled and engineered to handle vibration and extreme temperatures.

The company says its total computing performance has increased twentyfold in eight years, based on experience from more than 200 million fully autonomous miles. It also works with AMD, Micron, Nvidia, Samsung, Sandisk, Socionext and TSMC on different parts of the platform.

Building custom silicon gives Waymo tighter control over cost, power use and latency as it expands robotaxi service. It also shows how autonomous-vehicle companies are becoming chip designers, not just software developers. Whether that vertical integration produces safer and more economical operations will depend on real-world results across the cities Waymo enters.

See an error? Read our corrections policy or email [email protected].

MORE FROM TECHNOMALIST

Continue reading

View all
Telecommunications network equipment representing online safety controls
Internet

PLDT Moves to Block Over 100 Sites in Child Online-Safety Push

Remote business meeting shown across multiple computer screens
Software

Microsoft Teams Adds Automatic Blocking for External Meeting Bots

Identity security dashboard representing a blocked cyberattack
Cybersecurity

ReliaQuest Says Device Trust Stopped ShinyHunters Data Theft