Waymo has quietly stopped buying Nvidia chips for its robotaxis. The Alphabet subsidiary developed its own silicon in-house, and it’s already running in the company’s newest vehicles, according to Bloomberg. The chip crunches more than 1,000 TOPS — that’s trillions of operations per second — and TSMC fabricates it on a 5-nanometer node. That’s the same ballpark as Nvidia’s current autonomous driving hardware, which tells you Waymo isn’t compromising on compute to save a buck.
The stated goal is cost reduction, which has been the stubborn elephant in the room for every self-driving company since the first lidar-equipped prototypes hit public roads a decade ago. Waymo says the custom chip processes sensor data faster and runs the AI models that help its vehicles interpret what’s happening around them. Until now, the company mixed Nvidia and AMD parts to get the job done. Bringing that design in-house means one less supplier to negotiate with — and a lot less margin going to Jensen Huang.
This move lands just as Waymo starts building its next robotaxi with Zeekr, the Chinese EV maker. The new chip goes into that vehicle. It’s a notable pairing: American-designed AI silicon married to Chinese-manufactured hardware, at a moment when both governments are scrutinizing exactly that kind of cross-border technology flow. Waymo has been scaling aggressively in San Francisco, Phoenix, and Los Angeles, and the unit economics of every ride depend heavily on what’s under the hood.
For Nvidia, this is a slow burn rather than a splashy loss. Waymo’s vehicle fleet is modest compared to the millions of cars Nvidia’s DRIVE platform targets across the broader automotive industry. But the signal is unmistakable: the most sophisticated autonomy players are willing to spend years and deep engineering resources to own their silicon stack. Tesla did it first with its in-house FSD chip. Now Waymo. The question is whether the rest of the industry follows, or whether Nvidia’s software ecosystem keeps them locked in.
