Waymo reveals custom 5nm robotaxi chip delivering over 1,000 TOPS

Custom silicon under the robotaxi

Waymo has been expanding its robotaxi service city by city, and this week the company explained a key piece of hardware behind that push. According to TechCrunch, Waymo built a custom 5-nanometer ASIC designed to handle the flood of raw data produced by its vehicles before that data reaches the core "brain" of the self-driving system.

The chip is part of Waymo's sixth-generation self-driving system, which debuted in the company's next-generation robotaxi, the Ojai. That vehicle carries 13 high-fidelity cameras, and Waymo says the new silicon delivers more than 1,000 TOPS — trillions of operations per second — of computing performance. As TechCrunch points out, that figure puts the chip in roughly the same performance range as Nvidia's DRIVE AGX Thor, a processor built for automated driving applications.

A cost story as much as a performance story

The disclosure came in the same week that Waymo opened the Ojai to all riders in Los Angeles, Phoenix and San Francisco, per TechCrunch. The company has repeatedly framed the vehicle as cheaper to build, operate and maintain than earlier hardware generations — economics it considers essential to any future profitability. The custom chip fits squarely into that narrative: designing silicon in-house gives Waymo tighter control over the cost and efficiency of the compute inside every vehicle it deploys.

Waymo argues that the chip is a critical piece of a system able to react quickly and safely in complex, high-density environments such as city streets.

Vertical integration, with a deep bench

The news also clarifies just how hands-on and vertically integrated Waymo has become while chasing that goal, TechCrunch notes. But the company is not doing all of this alone. Waymo published a list of compute partners — some named for the first time — that includes AMD, Micron, Nvidia, Samsung, Sandisk, Socionext and TSMC. In practice, that means Waymo owns the chip design while relying on a broad semiconductor supply chain for manufacturing and components.

The Nvidia relationship is particularly interesting. Nvidia is listed among Waymo's partners, yet Waymo's in-house ASIC lands in the same performance territory as Nvidia's flagship automotive processor — making the chipmaker simultaneously a supplier and, at the compute level, a competitor.

Why it matters

Custom silicon is expensive and slow to develop, so a company committing to its own 5nm ASIC is signaling that compute — not just software and sensor suites — is now a core competitive lever in autonomous driving. If Waymo can efficiently process the output of 13 high-fidelity cameras at a lower cost per vehicle, that directly improves the unit economics of every robotaxi it puts on the road.

It also shifts the industry's balance of power. Autonomous vehicle developers have often leaned on off-the-shelf compute platforms, and Waymo's move suggests the largest players intend to own more of the stack, from driving software down to the chip itself. That carries implications for suppliers racing to become the default compute platform for automated driving, and it raises the bar for rivals deciding whether to buy their silicon or build it.

For riders, the more immediate effect is operational: cheaper, more efficient vehicles make it more plausible that Waymo's rapid, city-by-city expansion can keep going — and, eventually, pay for itself.