Waymo’s 5nm custom chip is the clearest sign yet that the robotaxi company has quietly become a serious hardware developer, not just a software-and-sensor shop bolted onto someone else’s compute stack.
The disclosure, published on the Waymo blog and co-authored by the company’s vice-president of engineering and its compute lead, confirms that Waymo built a bespoke application-specific integrated circuit (ASIC) fabricated on the TSMC 5nm automotive node. The chip delivers more than 1,000 TOPS (trillions of operations per second) of computing performance, putting it roughly in the same performance range as Nvidia’s DRIVE AGX Thor processor.
Inside the Waymo 5nm Custom Chip
The Waymo 5nm custom chip handles front-end sensor processing only. Specifically, it performs tasks such as temporal denoising to sharpen perception in low-light conditions, then feeds processed data into a separate inference engine running sensor-fusion models, according to The Next Web. Everything else (orchestration, data movement, logging) runs on components supplied by AMD, Micron, Samsung, SanDisk, and Nvidia.
That division of labour matters. The chip is not a replacement for Waymo’s partner ecosystem; it is a targeted insertion designed to reduce latency at the point where raw sensor data enters the system. The Ojai robotaxi carries 13 high-fidelity cameras, and that front-end data volume is precisely the bottleneck the ASIC is meant to clear.
Waymo describes the broader compute architecture as built around three principles it calls ‘responsive, ruggedised, and redundant,’ and says the ASIC is ‘just one of several exciting custom components we’re developing,’ with more in-house silicon planned. The TSMC automotive node on which the chip is fabricated completed automotive-grade qualification in 2022 and entered volume production for automotive customers in 2025, making Waymo’s deployment one of the earliest large-scale commercial uses of that node, per TSMC’s 2025 annual report as cited by TechTimes.
For context on where this sits competitively: Tesla’s current AI4 chip runs the full FSD stack including planning. Waymo’s ASIC handles only the front-end layer. Tesla’s next-generation AI5 chip taped out in April 2026 and is expected to deliver roughly five times the compute of AI4, with vehicle integration anticipated in 2027. The two companies are solving related but structurally different problems with silicon.
Waymo Compute Lead Daniel Rosenband is scheduled to deliver a keynote titled ‘Compute in Motion: Challenges of Autonomous Driving’ at Hot Chips 2026 at Stanford on 24 August, where additional technical detail is expected.
The scale context helps explain why the company is investing in custom silicon at all. As of March 2026, Waymo had accumulated 220.6 million rider-only miles and surpassed 20 million lifetime paid trips, logging approximately 500,000 weekly paid trips across a fleet of roughly 3,500 vehicles as of July 2026, according to Gadgets Now. Its sixth-generation Driver cut sensor count by 42 per cent versus the prior generation. Custom silicon is one way to maintain performance whilst trimming cost per vehicle.
Also Raises $150 Million, and the Lidar Question Nobody Will Answer
Elsewhere in mobility funding, Also, the startup incubated within Rivian and chaired by Rivian CEO RJ Scaringe, announced a $150 million Series D on 19 August 2026, led by Prysm Capital with participation from Eclipse, Greenoaks, and MVP Ventures. The snippet reports $455 million raised since Also spun out of Rivian in March 2025; eletric-vehicles.com tallies cumulative funding at approximately $655 million when including earlier tranches, though TechCrunch and Yahoo Finance report $455 million, suggesting a difference in how the initial spin-out capital is counted. The BusinessWire press release is the primary record.
The company has shifted from describing itself as a micromobility firm to a builder of ‘the world’s most capable driven and autonomous small electric vehicles.’ That pivot comes with a multi-year commercial agreement with DoorDash for autonomous delivery vehicles, announced alongside a $200 million round in early 2026, plus a separate deal with Amazon, announced in October 2025, to supply pedal-assist e-cargo quads for delivery fleets in Europe and the United States, according to Tech Funding News.
On the lidar front, the Idaho National Laboratory is evaluating whether Chinese lidar sensors pose a security risk if widely deployed on US vehicles, with funding reportedly coming from companies in the electric and autonomous vehicle industries. The lab told TechCrunch in both June and July that ‘this isn’t a project that we’re at liberty to discuss with reporters.’ Rivian, GM, Ford, Kodiak, Lucid, Nuro, and Uber all told TechCrunch they were unaware of the review; Aurora, Nvidia, and Zoox did not respond.
Hesai’s SEC filings disclose that the Chinese government ‘may influence or intervene in our operations at any time.’ The company claims its sensors store no data, citing a lack of onboard memory, though that addresses storage rather than transmission. A decade ago, lidar sensors cost as much as $75,000; Hesai and RoboSense have since cut prices to a few hundred dollars in some cases, which is part of what makes the supply-chain question more than academic.
Waymo’s Hot Chips keynote on 24 August is the next concrete moment to watch. If the company’s compute lead is presenting additional specifications to a room full of chip engineers, the gap between Waymo’s public narrative and its actual silicon roadmap is about to get a lot narrower.
