What changed
Alphabet's autonomous driving unit Waymo publicly disclosed a custom application-specific integrated circuit (ASIC) built on Taiwan Semiconductor Manufacturing Co.'s 5-nanometer process, marking the first time the company has detailed the onboard compute platform that powers its driverless vehicles. The chip is designed to ingest raw sensor data at the edge before handing information to the core driving system, rather than managing every vehicle maneuver. Waymo framed the silicon as part of a state-of-the-art system tailored to the in-vehicle operating domain with real-time requirements.
The measured result
The purpose-built ASIC runs temporal denoising for low-light driving and performs sensor fusion across 13 high-resolution cameras, four lidars and radar feeds, collectively delivering more than 1,000 trillion operations per second (TOPS) of machine-learning compute. Waymo said onboard compute capacity has risen 20-fold over the past eight years, and that the system is engineered for responsiveness, ruggedness and redundancy, with redundant computers able to take over if one experiences a fault, since no human driver is available to intervene. The chip is already entering production in the Ojai, Waymo's purpose-built passenger vehicle developed with Geely-owned Zeekr.
Hybrid architecture and supplier ties
Rather than displacing third-party chipmakers, Waymo is adopting a hybrid computing architecture that pairs its in-house machine-learning silicon with externally sourced CPUs, GPUs, accelerators, memory and storage. Confirmed hardware partners include AMD, Micron, Nvidia, Samsung Electronics, SanDisk, Socionext and TSMC, with Waymo handling orchestration, data movement and logging across the heterogeneous system. Contract sizes, part allocations and expected purchase volumes were not disclosed, leaving the near-term revenue impact for each supplier unclear from this announcement.
Fleet context and rollout
The disclosure comes as Waymo expands commercial robotaxi services in Phoenix, San Francisco and Los Angeles, reportedly handling roughly 500,000 paid trips weekly, while pushing to scale production of its next-generation driver system. The sixth-generation system simplifies the sensor configuration and shifts more processing to custom chips, with the aim of reducing hardware complexity and cost while preserving multiple sensing modalities, and is intended to be applied across multiple vehicle platforms beyond Ojai.
What remains uncertain
Waymo has not specified which components each supplier provides, the size of related contracts or expected purchase volumes, making it difficult to estimate the announcement's effect on any individual chipmaker's results. It is also unclear when the custom silicon will reach higher-volume deployment across the broader fleet, and whether the hybrid architecture will shift toward greater in-house design over time as the Ojai rollout progresses.
Waymo's next verifiable milestone is the production deployment of the Ojai robotaxi with the custom 5nm ASIC integrated, alongside continued expansion of paid commercial service in its existing U.S. cities.
Share this article
