Waymo has pulled back the curtain on the computing architecture that keeps its driverless vehicles safe, revealing a dual-computer system designed to handle failures without waiting for human intervention from miles away. The company disclosed that its robotaxis rely entirely on onboard computing to make driving decisions in milliseconds, with multiple layers of backup hardware ensuring the vehicle can reach a safe stop if something goes wrong.

What Happens When a Waymo's Main Computer Fails?

For passengers settling into the back seat of a driverless Waymo, the empty driver's seat raises an obvious question: what if the computer doing all the driving suddenly stops working? Waymo's answer involves redundancy built into every critical system. The company describes its newest compute system as working somewhat like two independent engines that can run full workloads in parallel during normal driving.

If one computing system develops a fault, the other takes over immediately. This approach eliminates the need to wait for a remote employee to intervene. Waymo's safety documentation indicates that a secondary onboard computer can bring a vehicle to a safe stop if it detects a failure in the primary system. Beyond computing redundancy, the vehicles also feature backup braking and redundant steering hardware, with separate power sources supporting critical driving systems.

How Does Waymo Process Driving Decisions So Quickly?

Waymo's vehicles take in enormous amounts of sensor data every second. The company's latest sixth-generation sensor system uses 13 cameras along with lidar and radar sensors to build a detailed picture of the road ahead. All of this information flows into an onboard computer that must make sense of it fast enough to steer, brake, or react to hazards entering the road.

The critical point is that these driving decisions happen entirely onboard. The vehicle does not need to send data to a distant data center before making a turn. According to Waymo, it has increased its raw computing power approximately 20 times over the past eight years, which helps shrink the time between seeing something and reacting to it.

To handle this workload, Waymo revealed a custom 5-nanometer ASIC, which stands for application-specific integrated circuit. Think of it as a chip designed specifically for the job of processing what the car sees, rather than a general-purpose processor asked to do everything. The chip handles raw information flowing in from the vehicle's sensors and helps process camera, lidar, and radar data before that information reaches other parts of the driving system.

Waymo says the ASIC delivers more than 1,000 TOPS of machine-learning performance, where TOPS means trillions of operations per second. In practical terms, the system can process footage from 13 high-resolution cameras simultaneously and improves what the cameras can detect in low light conditions.

Steps to Understanding Waymo's Redundancy Architecture

• Dual Computing Systems: Two independent computers run full workloads in parallel during normal driving, allowing one to take over instantly if the other fails without requiring remote assistance.
• Hardware Backups: Secondary braking and steering systems with redundant controllers and power supplies ensure the vehicle can stop safely even if primary hardware fails.
• Independent Power Sources: Critical driving systems have separate power supplies so a single electrical problem does not automatically disable all vehicle functions.
• Continuous Self-Monitoring: Waymo's vehicles continually check their own systems while driving, detecting problems before they become safety issues.

What Role Do Remote Operators Actually Play?

A common misconception about driverless robotaxis is that employees sit in offices with steering wheels and pedals, ready to take control remotely. Waymo clarified that this is not how its system works. The company does use what it calls Remote Assistance, but those employees provide information rather than remotely driving the vehicle.

If the Waymo Driver encounters an unusual situation, it can request additional context from a remote agent. The car then decides whether to use that information. Waymo emphasized that remote agents do not continuously watch every vehicle. The automated driving system remains responsible for controlling the car, so if the onboard computer develops a hardware fault, a remote employee does not suddenly become your driver.

Can Redundancy Prevent All Problems?

While backup computers protect against hardware failures, they cannot prevent every problem a driverless vehicle may encounter. A computer can be functioning exactly as designed while the software struggles to interpret an unusual road situation. Waymo itself experienced this when it issued a software recall involving 3,871 vehicles after robotaxis entered closed freeway construction zones in Phoenix and the San Francisco Bay Area.

That problem involved how the automated driving system handled changing road conditions rather than a dead onboard computer. It serves as a useful reminder that redundancy addresses one category of risk. Autonomous driving also depends on perception, decision-making, and the software correctly understanding what is happening outside the car.

Waymo's latest published safety analysis covered more than 220 million fully autonomous miles through March 2026. According to the company's own safety research, its vehicles experienced fewer serious or fatal injury crashes than human drivers operating in comparable areas during that analysis period.