Waymo develops its own chip for its robotaxis, reducing its reliance on Nvidia

Waymo, Alphabet's autonomous driving division, has developed a proprietary chip for its robotaxis. The application-specific integrated circuit (ASIC) handles front-end sensor processing on board the vehicles and is already in use across the current fleet. With this move, Waymo is partially stepping away from its earlier dependence on chipmakers such as Nvidia for this part of the computing stack.

The chip delivers more than 1,000 TOPS (trillions of operations per second) of machine learning compute power, comparable to what Nvidia offers in its latest autonomous driving systems. The ASIC combines raw data from thirteen high-resolution cameras with signals from radar and lidar sensors in real time, before that information is passed to Waymo's driving software. Production is handled by TSMC, which applies its 5-nanometre process.

Waymo shared the details in a blog post signed by Vice President of Engineering Satish Jeyachandran and Compute Lead Daniel Rosenband. Further technical explanation will follow at the Hot Chips conference at Stanford's Memorial Auditorium in Palo Alto.

What the chip does and why Waymo builds it in-house

Front-end processing and fusion of sensor streams required a specialised chip. General-purpose processors from external suppliers are designed for broad applications; Waymo's own ASIC is tailored to precisely this one set of tasks within the driving system.

The design is based on more than 200 million fully autonomously driven kilometres. That real-world data has informed which characteristics the chip needed to have. Alongside compute performance, physical requirements also apply: robustness against vibrations, shocks and extreme temperatures. The chip also features two independent engines running in parallel, providing redundancy should one of the two fail.

The development team consists largely of former engineers from Intel and Qualcomm. For the broader computing stack, Waymo continues to work with other partners, including AMD, Micron, Nvidia, Samsung, Sandisk, Socionext and TSMC. The ASIC therefore does not replace Nvidia entirely; it specifically concerns the front-end sensor processing layer.

Waymo's current position and scale

Waymo provides more than 500,000 fully autonomous rides per week using electric vehicles. The company is active in eleven US cities and has plans to expand to nearly twenty additional cities. London and Tokyo are also on the expansion list. The total compute power used by the system has increased twentyfold over the past eight years.

The new robotaxi for which the chip was partly designed is being built in collaboration with Chinese automotive brand Zeekr. In February 2026, Waymo closed a $16 billion funding round at a valuation of $126 billion. Participating investors included Sequoia Capital, Andreessen Horowitz, DST Global, Dragoneer Investment Group, Mubadala and parent company Alphabet.

The company also faced two recalls this year. In May, nearly 3,800 vehicles were recalled following a software issue that caused cars to drive through flooded roads; in June, a recall of nearly 3,900 vehicles followed due to a fault whereby vehicles could enter closed highway construction zones.

A broader move towards self-designed chips

With this step, Waymo is following a pattern already visible at several large technology companies. Apple, Google and Amazon have been designing their own chips for specific tasks in their products and data centres for years, rather than relying entirely on chipmakers such as Intel or Nvidia. Tesla has long built its own AI training and inference hardware. Waymo now doing the same for sensor processing in its vehicles fits within that trend.

The benefits are twofold. A self-designed chip can be more precisely tuned to the specific computing task, which can be more efficient than a general-purpose solution. At the same time, it reduces direct dependence on a single external supplier for a critical component of the system. Waymo itself emphasises that Nvidia and other partners remain on board for the rest of the stack.

For the European and Dutch AI and mobility sector, this development is a signal that vertical integration of hardware and software is becoming an increasingly prominent part of the strategy of major autonomous driving players. European companies and policymakers working on their own AV programmes or AI hardware policy can observe here how a major market participant is choosing greater control over the full technology stack, from sensor to driving software.