Key Points:
- Waymo officially introduced its first custom-designed 5-nanometer application-specific integrated circuit chip to power next-generation robotaxis.
- The specialized processor delivers over 1,000 TOPS of machine learning compute to clean and fuse raw camera, lidar, and radar data in real time.
- The chip is currently in production and deployed inside Waymo’s new purpose-built Ojai robotaxi fleet manufactured with Zeekr.
- Waymo revealed a heterogeneous computing architecture, partnering with seven semiconductor leaders, including Nvidia, AMD, TSMC, and Micron.
Alphabet’s autonomous driving subsidiary Waymo has unveiled its first custom-designed computer chip created specifically for its commercial robotaxi fleet. The purpose-built application-specific integrated circuit (ASIC) marks a major turning point in the company’s hardware engineering strategy, giving the autonomous vehicle leader direct control over its onboard computing stack as it scales driverless ride-hailing services across major American metropolitan markets.
Manufactured on an advanced 5-nanometer process node by Taiwan Semiconductor Manufacturing Company, the custom processor delivers over 1,000 trillion operations per second (TOPS) of dedicated machine learning compute performance. This computational power matches the performance benchmarks of top-tier commercial autonomous driving platforms, providing the high-speed processing needed to run complex neural networks directly onboard without relying on remote cloud connectivity.
The custom silicon serves a specialized operational role as a front-end processing engine. Rather than attempting to handle every non-driving computational task, the chip focuses exclusively on ingesting the massive flood of raw sensory inputs coming from external high-resolution cameras, imaging radar, and lidar arrays. The processor performs instant temporal denoising, filters out optical noise during low-light and severe weather conditions, and combines multi-modal sensor streams before routing clean information to the vehicle’s core artificial intelligence driving brain.
The custom chip is already rolling off production lines and operating inside Waymo’s latest sixth-generation Driver architecture. The hardware is actively deployed inside the company’s new purpose-built Ojai robotaxi platform, manufactured in partnership with electric vehicle brand Zeekr and outfitted with autonomous systems at an integration facility in Mesa, Arizona. Passengers hailing rides through the company’s mobile application in Los Angeles, Phoenix, and San Francisco are already beginning to experience the upgraded vehicle platform on public roads.
Company engineers clarified that designing proprietary silicon does not mean abandoning established commercial semiconductor suppliers. Instead, the autonomous vehicle maker introduced a balanced, heterogeneous computing architecture that pairs its custom front-end machine learning chip with high-performance processors, graphics chips, and memory solutions from seven named industry leaders: AMD, Micron Technology, Nvidia, Samsung Electronics, SanDisk, Socionext, and TSMC.
The custom silicon rollout arrives as Waymo solidifies its commanding lead in commercial autonomous ride-hailing. The company operates a fleet of approximately 4,000 autonomous vehicles across more than 10 major metropolitan areas, completing over 500,000 paid passenger trips every week. With more than 200 million fully autonomous commercial miles logged on public roads, the enterprise has accumulated unmatched real-world driving data to guide its hardware optimizations.
Developing in-house silicon allows the enterprise to overcome critical engineering bottlenecks that challenge generic automotive chips. By co-designing hardware, proprietary sensors, and machine learning models side by side, engineers can optimize memory bandwidth, reduce computing latency from sensor to steering action to mere milliseconds, and significantly lower total vehicle hardware costs. Lowering per-vehicle compute costs is essential for achieving long-term commercial profitability as fleet sizes expand into tens of thousands of vehicles.
The custom autonomous driving chip aligns with a broader corporate strategy across parent company Alphabet to design proprietary silicon for intensive artificial intelligence workloads. From Tensor Processing Units in hyperscale data centers to custom processors in smartphones, building purpose-fit chips reduces dependence on third-party pricing power and protects profit margins across both digital and physical computing operations.
As rival autonomous vehicle programs ramp up testing in major cities, Waymo’s vertical hardware integration sets a new technical benchmark for the robotaxi sector. By combining custom 5nm silicon with proven sensor fusion and commercial fleet scale, the autonomous pioneer establishes a reliable computing foundation to power the expansion of driverless mobility into new metropolitan markets across the United States.
