Waymo Builds Custom Robotaxi Chip to Reduce Nvidia Reliance

Alphabet's self-driving subsidiary, Waymo, has designed a custom processor to power its autonomous vehicles, marking a significant shift away from off-the-shelf hardware. The new silicon is already deployed in the company's latest generation of robotaxis. According to reports, the custom chip is manufactured by Taiwan Semiconductor Manufacturing Company (TSMC) using a 5-nanometer process node.

The custom hardware delivers more than 1,000 trillions of operations per second (TOPS) of performance. This processing capability puts the chip on par with Nvidia's current autonomous driving platforms. Waymo designed the chip specifically to accelerate the processing of complex sensor data and run the deep learning models that allow its vehicles to perceive and navigate real-world environments. Previously, the company relied on a combination of processors from Nvidia and AMD to handle these intensive computational workloads.

Waymo is integrating this new silicon into its next-generation robotaxis, which are being co-developed with the Chinese electric vehicle manufacturer Zeekr. By tailoring the hardware to its specific software stack, Waymo aims to optimize performance while significantly reducing the overall production costs of its autonomous fleet.

For autonomous vehicle engineers and industry practitioners, Waymo's transition to custom silicon highlights a growing trend of vertical integration among leading AI firms. By bypassing general-purpose chips, developers can eliminate computational bottlenecks and achieve better thermal and energy efficiency. This milestone demonstrates that custom-designed hardware is becoming essential for scaling robotaxi operations economically, potentially forcing other autonomous vehicle developers to reconsider their reliance on standard merchant silicon.