Rivian unveils Gen 3 autonomy computer built on Arm’s Armv9 platform
At its Autonomy & AI Day in Palo Alto on Dec. 11, Rivian announced a third‑generation on‑board computer designed to run the company’s “physical AI” stack. The new system, dubbed the Gen 3 Autonomy Computer, is powered by a custom‑designed silicon package called the Rivian Autonomy Processor 1 (RAP1). Rivian said the chip was co‑engineered with Arm Ltd. and fabricated on a 5‑nm process at TSMC.
According to a TechCrunch report, RAP1 delivers 1,600 INT8 TOPS of inference performance and can process roughly 5 billion pixels per second—four times the capability of Rivian’s current Nvidia‑based system. The processor integrates two Arm Cortex‑A720AE cores, Arm’s most recent Armv9‑based CPUs, along with a bespoke accelerator for sparse tensor operations.
How the RAP1 chip leverages Armv9 to power edge‑centric autonomy
Armv9, introduced in late 2023, adds a suite of security and performance extensions that are attractive for safety‑critical workloads. The architecture includes Pointer Authentication (PAC) and Branch Target Identification (BTI), features that help protect against memory‑corruption attacks—a prerequisite for any system that makes real‑time driving decisions.
Rivian’s engineering lead, Wassym Bensaid, explained that the Cortex‑A720AE cores handle high‑level perception and planning while the on‑die accelerator runs the dense convolutional neural networks that predict vehicle trajectories. Because Arm’s IP is designed for low power consumption, Rivian can sustain the compute load without eroding the electric range of its vehicles, an issue that has plagued early attempts at on‑vehicle AI.
Business implications and the broader physical‑AI market
The announcement arrives as analysts estimate the “physical AI” market—covering autonomous vehicles, robotics, and edge computing—to exceed $2 trillion by 2030, according to a Bloomberg analysis. Rivian’s move to in‑house silicon mirrors strategies at Tesla and Waymo, which have long pursued custom chips to reduce dependence on external suppliers and improve cost margins.
Rivian plans to roll out the new hardware on its upcoming R2 SUV later in 2026. The first‑generation R2 models will launch without RAP1 or lidar, but the company said future yearly updates will add “eyes‑off” and “personal Level 4” capabilities through a subscription known as Autonomy+. Pricing was disclosed as either a $2,500 one‑time fee or $49.99 per month, comparable to Tesla’s Full‑Self‑Driving add‑on.
The partnership with Arm also aligns with Rivian’s broader financing strategy. In 2024, the automaker secured a $6 billion joint‑venture investment from Volkswagen, which includes a technology‑sharing component that could accelerate the deployment of Arm‑based compute across both firms’ future models.
Expert views on safety‑centric compute for autonomous machines
Arm’s Executive Vice President of the Physical AI Business Unit, Drew Henry, said, “By combining Armv9’s security extensions with a high‑performance accelerator, we give automakers a trustworthy foundation that can evolve for years of on‑road operation.” The statement reflects a growing consensus among regulators that safety‑critical AI must be anchored in hardware that can be independently verified.
Industry analyst Todd Hester of TechInsights noted that Rivian’s choice of a dual‑core Arm design “balances the need for deterministic latency with the flexibility to run large language model‑style reasoning in the vehicle,” a capability that could enable future features such as natural‑language driver assistants.
While the hardware marks a clear technical step forward, the path to fully driver‑less operation still depends on extensive validation. The National Highway Traffic Safety Administration (NHTSA) recently tightened its guidelines for Level 4 systems, requiring documented fail‑safe mechanisms and over‑the‑air update security—areas where Arm’s built‑in safety architecture may prove advantageous.
What’s next for Rivian and the autonomy ecosystem?
Rivian’s next milestones include integrating the RAP1 into production R2 models in early 2026 and expanding the lidar sensor suite announced alongside the chip. The company also plans to open an API for third‑party developers to tap into its “Rivian Unified Intelligence” platform, which will run on the same Arm‑based compute stack.
For observers tracking the convergence of AI and mobility, Rivian’s announcement underscores the strategic shift toward vertically integrated silicon solutions. As Arm reports record‑high licensing revenue—$2.7 billion in FY 2024—more OEMs are likely to follow suit, seeking the performance‑per‑watt and built‑in safety features that are essential for scaling physical AI worldwide.
Read more on Globally Pulse’s technology coverage.
