Waymo Launches In-House Robotaxi Chip, Begins Mass Production with Nvidia-Comparable Performance

Alphabet's Waymo has built its own custom compute chip for robotaxis and moved it into production in its latest vehicle generation, a step toward vertical integration of the self-driving stack. The application-specific integrated circuit runs above 1,000 TOPS, which puts it level with Nvidia's latest autonomous driving platforms on that particular measure, and is manufactured on Taiwan Semiconductor's 5-nanometer process.

The chip arrives alongside Waymo's newest robotaxi platform, and the recently launched vehicle built with Zeekr, an arm of China's Geely, is being equipped with it. Waymo continues to source other components from outside vendors, so this is supply diversification rather than a clean break; the practical effect is to reduce dependence on a single external compute supplier and to allow hardware and software to be designed together.

For GOOGL the read is about unit economics. Autonomy is a cost problem before it is a demand problem, and compute is one of the largest bills of material in a driverless vehicle. Designing the silicon in-house lets Waymo target its own perception workload rather than pay for general-purpose headroom it does not use. For NVDA, one large autonomy customer designing its own part is a narrow revenue question, but it is the same pattern already visible in hyperscaler AI accelerators.

The claim worth testing is the performance parity: a TOPS number is a peak throughput figure, not a measure of real-world latency, thermal behavior, or energy efficiency across driving conditions, and no independent benchmark exists. Watch for fleet-level reliability data, the pace at which the new chip propagates across Waymo's vehicle generations, and whether third-party sourcing actually declines over time.

Related Stocks

• GOOGL
• NVDA
• AMD