TechCrunch Mobility: The hidden human cost of robotaxis

One important clarification on what I wrote about last week. You might recall that Waymo shared information about its custom silicon chip — specifically a 5 nm ASIC chip that is designed to handle the massive influx of raw data before it reaches the core “brain” of the self-driving system. Waymo stated in its blog post:

“While these ASICs alone deliver over 1,000 TOPS of ML performance dedicated to front-end processing and ML models, we optimize across the full stack to maximize achieved performance, especially in the low-batch regimes we often operate.”

That line led me, and others, to believe that its one chip can deliver 1,000 TOPs (trillions of operations per second) of computing performance. I compared it to Nvidia’s Drive AGX Thor automotive processor, noting it was about the same performance. 

But alas, one eagle-eyed reader reached out with some questions, which prompted me to turn to Waymo for official answers. A Waymo spokesperson told me, “It’s for the system, not a single chip.”

That’s an important distinction.