Home · Science & Technology
Waymo details custom 5nm chips used for robotaxi sensor fusion
Confirmed
What we know
Alphabet's Waymo published a technical look at the compute stack in its robotaxis, introducing a purpose-built 5-nanometer ASIC aimed at cleaning, fusing, and running neural networks on raw camera, lidar, and radar data before the core driving stack.
Waymo said the ASICs alone deliver over 1,000 TOPS of machine-learning performance for front-end processing. It listed partners including AMD, Micron, NVIDIA, Samsung, Sandisk, Socionext, and TSMC, and pointed to upcoming Hot Chips talks.
What's confirmed
- Waymo's Aug 2026 engineering blog describes the custom 5nm ASIC role in sensor fusion and front-end ML, and states ASICs deliver over 1,000 TOPS for that front-end work.
- Waymo named suppliers AMD, Micron, NVIDIA, Samsung, Sandisk, Socionext, and TSMC.
- Independent tech outlets (The Verge, TechRepublic and others) reported the disclosure and summarized the architecture as heterogeneous rather than a single all-in-one driving SoC.
What's still developing
- Waymo has not published a full public bill of materials with chip counts, power draw, and end-to-end latency numbers in the blog post; further detail may come at Hot Chips.
- How the custom silicon changes unit economics as Waymo scales fleets remains a forward-looking question for analysts, not a measured company metric in the disclosure.