Waymo puts a custom 5 nm chip in its robotaxi system
Thirteen high-fidelity cameras feed Waymo’s next-generation Ojai robotaxi. Behind that stream, the company has built a custom 5 nm ASIC — a chip designed for a specific job — to process raw sensor data before it reaches the vehicle’s central self-driving computer, TechCrunch reports. The disclosure arrives as Waymo’s Ojai robotaxi opens to all riders in Los Angeles, Phoenix and San Francisco.
Waymo says the chip delivers more than 1,000 TOPS, or trillions of operations per second. That puts it roughly in the same performance range as Nvidia’s DRIVE AGX Thor, an automotive processor designed for automated driving. The custom chip is not the whole system: its role is to handle the first surge of data so the core “brain” can work with information arriving from the vehicle’s sensors.
The engineering choice is tied to a practical constraint. Waymo has repeatedly said the next-generation vehicle is cheaper to build, operate and maintain — conditions the company considers necessary if its robotaxi business is eventually to become profitable. It says the chip is a critical piece of a system intended to react quickly and safely in dense city environments, but it gives no dollar figure for the savings.
So what changes in practice? A robotaxi operator trying to expand across cities gets a computing system designed around its own vehicles and sensor load, rather than relying entirely on off-the-shelf hardware. That could support Waymo’s push beyond existing service areas, with Waymo’s Ojai robotaxi open to all riders in three US cities. The immediate benefit is therefore operational: the custom silicon is part of a vehicle Waymo says is less costly to run and maintain.
The claim still has boundaries. The performance figures and efficiency claims come from Waymo, while TechCrunch describes the comparison with Nvidia’s processor as approximate; the report provides no independent benchmark, quantified cost reduction or evidence that the chip alone improves safety. Waymo is also not building the computing stack alone: its listed partners include AMD, Micron, Nvidia, Samsung, Sandisk, Socionext and TSMC.
Comments
Loading the thread…
Sign in to leave a comment. Sign in