Waymo has lifted the lid on the computing hardware that powers its robotaxis, revealing a custom-built silicon chip and detailed specifications of the "brain" that processes sensor data in real time. The company, which operates around 4,000 vehicles across more than 10 cities and conducts approximately 500,000 paid trips weekly, disclosed these technical details in a blog post, marking the first public glimpse into the heavy-duty computer systems housed in vehicle trunks.

What Makes Waymo's Robotaxi Computer Different?

Unlike traditional vehicles, Waymo's robotaxis rely on a sophisticated computing stack that processes massive streams of data from cameras, lidar sensors, and radar systems. The company's custom 5-nanometer application-specific integrated circuit (ASIC) sits at the front lines of this system, cleaning and combining sensor inputs while running artificial intelligence checks as data arrives. This front-end processor delivers 1,000 trillion operations per second (TOPS), enabling the vehicle to understand its surroundings and make driving decisions in milliseconds.

The design philosophy behind Waymo's compute system rests on three core principles.

"Compute systems for autonomous driving handle highly diverse workloads. At Waymo, we are designing a state-of-the-art system that would be considered impressive for a data center, with the added complexity of an in-vehicle operating domain and real-time requirements," stated Satish Jeyachandran, VP of Engineering, and Daniel Rosenband, Compute Lead.

Satish Jeyachandran, VP of Engineering, and Daniel Rosenband, Compute Lead, at Waymo

These three principles are non-negotiable for safe autonomous operation:

• Responsive: The system must process sensor data with minimal latency, requiring massive computing power that Waymo has scaled 20 times over the past eight years.
• Ruggedized: The hardware must withstand road vibrations, temperature extremes, and the physical demands of continuous operation in real-world conditions.
• Redundant: Critical systems must run in parallel, allowing the vehicle to continue operating safely if one component fails.

The computing stack also needs to fit within the vehicle's trunk without consuming all available space for passengers and luggage, operate nearly silently to preserve the rider experience, and maintain efficiency across diverse driving scenarios.

How Does Waymo's Sensor Fusion Strategy Compare to Tesla's Approach?

Waymo's multi-sensor architecture directly contrasts with Tesla CEO Elon Musk's dismissal of lidar and radar systems. Musk has called lidar a "crutch" and argued that combining data from multiple sensor types creates dangerous ambiguity when signals disagree. However, Waymo contends that its robust, redundant system is precisely what enables safe deployment at scale.

The company built what it describes as a machine learning-primary architecture that runs advanced neural networks at minimal latency. Rather than relying on a single sensor type, Waymo pairs its machine learning technologies with the best central processing units (CPUs), graphics processing units (GPUs), and specialized accelerators available. This balanced, heterogeneous approach allows the system to process high-fidelity data from 13 high-resolution cameras simultaneously and in real time, delivering improved low-light perception that traditional single-sensor systems often miss.

What Hardware Partners Power Waymo's Fleet?

Rather than building every component in-house, Waymo relies on a global supply chain of specialized manufacturers. The company partners with seven major hardware suppliers to assemble its cutting-edge compute system:

• AMD: Provides processors and computing architecture components for the system.
• Micron: Supplies memory and storage solutions for real-time data processing.
• Nvidia: Contributes graphics processing units and AI acceleration hardware.
• Samsung: Provides semiconductor components and memory technologies.
• SanDisk: Supplies storage solutions for the vehicle's computing needs.
• Socionext: Contributes specialized semiconductor designs.
• TSMC: Manufactures Waymo's custom 5-nanometer silicon chip.

This approach allows Waymo to focus its internal engineering efforts on real-time sensor fusion and front-end machine learning, the core competencies that differentiate its autonomous driving system.

What Are the Cost Implications of Waymo's Hardware Strategy?

The computing hardware represents a significant expense for Waymo's fleet expansion. Current estimates place the cost of sixth-generation hardware at $20,000 to $25,000 per vehicle, a substantial decrease from the fifth-generation's estimated $100,000 to $125,000 per vehicle. Despite this dramatic cost reduction, hardware expenses remain a potential constraint on Waymo's ability to scale profitably.

The company has demonstrated its commitment to cost optimization through successive hardware generations, but the challenge of building a profitable business model remains distinct from the engineering achievement of building a capable autonomous vehicle. As Waymo continues expanding its fleet, managing hardware costs while maintaining safety and performance standards will be critical to long-term viability.

How Does the New Ojai Vehicle Perform in Real-World Conditions?

Waymo recently opened its new Ojai robotaxi to the public in the San Francisco Bay Area, Los Angeles, and Phoenix. The Ojai, sourced from Geely's Zeekr division and manufactured in China, represents a purpose-built passenger vehicle rather than a converted luxury car. The vehicle features two giant sliding doors with no center pillar, a flat floor, and generous interior space that makes entry and exit straightforward.

The Ojai includes three interior screens that can display route information, a visualization of what the car's sensors perceive, or entertainment and climate controls. The vehicle offers wireless and wired phone charging, comfortable seating for multiple passengers, and excellent visibility through large windows. However, early riders have noted some trade-offs compared to Waymo's previous Jaguar i-Pace vehicles.

Some users reported that the Ojai's ride quality felt rougher than expected, despite the advantage of a flat floor typical of electric vehicles. The vehicle's suspension does not yet incorporate dynamic adjustment based on road conditions detected by lidar sensors, a capability that could smooth the ride experience. Additionally, the Ojai is currently restricted from freeway operation, which extends trip times and limits the service area's practical utility for longer journeys.

The vehicle's design prioritizes functionality over aesthetics, with a boxy shape that maximizes interior space but may sacrifice aerodynamic efficiency. The trunk is taller but narrower than the i-Pace, potentially creating challenges for passengers with large amounts of luggage. Despite these limitations, the Ojai demonstrates Waymo's commitment to designing vehicles specifically for autonomous operation rather than retrofitting existing platforms.