Most folks would think of Rivian as a runner-up to Tesla thanks to its aluminum unibody R1 vehicles, but the automaker has actually been making a tidy sum in software and services. True to that statement, Rivian today announced its Rivian Autonomy Processor chip and Autonomy Compute Module 3 platform, ready for integration in its own vehicles, and likely those of other automakers.
The Rivian Autonomy Processor (RAP1) is a bespoke in-house effort. The chip itself is an Armv9 design, with 14 Cortex-A720AE cores. and is manufactured on a 5 nm process. It has support for the company's own RivLink (no relation or comparison with Nvidia's NVLink), an interconnect that Rivian says provides extensibility to computing power by adding other chips, presumably similar, but there aren't many details.
The RAP1 is the cornerstone of the Autonomy Compute Module 3 (ACM3), a vehicle computer system focused on self-driving capabilities. Its claimed abilities include 1,800 TOPS of INT8 inference, enough grunt to process 5 billion pixels per second of camera feeds.
Interestingly, Rivian says that ACM3 will support the addition of LiDAR input, something that Tesla decided against and Elon Musk himself derided. Rivian says the ACM3 and LiDAR are undergoing validation and will be used in versions of its upcoming R2 vehicle come late 2026.
Existing second-generation R1 vehicles will get Universal Hands Free (UHF), a limited driving assistant akin to Tesla's Autopilot. Rivian says it'll be able to function for extended periods of time and covers 3.5 million miles of roads in the U.S. and Canada. UHF isn't limited to highways, either, as it's meant to work on roads that have clearly painted lines.
For those wishing for additional capabilities, Rivian will launch its Autonomy+ service, priced as a $2,500 one-time fee or as a $49.99/mo subscription. Probably wisely, the company chose its words carefully and didn't imply the vehicle would have complete self-driving ability, opting instead to tout Autonomy+ as a safety improvement, though it states it's on the race for Level 4 autonomous driving.
The company also talked up the Large Driving Model, its driving-oriented AI model (analogous to chatbots' Large Language Model) that Rivian claims is superior thanks to its use of Group Relative Policy Optimization, similar to how DeepSeek operates.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter Get Tom's Hardware's best news and in-depth reviews, straight to your inbox. Contact me with news and offers from other Future brands Receive email from us on behalf of our trusted partners or sponsors
Comparisons to Tesla vehicles' HW-series platforms are inevitable, especially given that recently, Elon Musk touted his auto business's acumen in the chip department. Given that Rivian is looking into a 2026 timeframe for RAP1, we can only compare via rumors of Tesla's upcoming HW5. Even still, with the appropriate amount of salt, Tesla is reportedly gunning for 2,000 to 2,500 TOPS for HW5, though naturally, there's no support for LiDAR.
Rivian's latest move might seem odd given it's primarily an automotive outfit, and there's no shortage of discussions about its upcoming $45,000 Rivian R2 SUV. However, the company signed a $5.8 billion joint venture with the VW Group in 2024 for EV software and electrical architecture.
... continue reading