Rivian Unveils AI Chip for Automated Driving, Ditches Nvidia
Automaker Will Equip Upcoming R2 SUVs With Rivian Autonomy Processor 1 Chips and New Lidar Sensor
Bloomberg News
[Stay on top of transportation news: .]
Rivian Automotive Inc. has developed its own artificial intelligence chip, replacing Nvidia Corp. technology as part of a broader push to add and enhance automated-driving features in future vehicles.
The automaker will equip its upcoming R2 sport utility vehicles with Rivian Autonomy Processor 1 chips and a new lidar sensor. Taiwan Semiconductor Manufacturing Co. will produce the chips that, combined with the new sensor and AI model developments, will bolster Rivian’s efforts to eventually offer autonomous driving capability.
“This is not a bet one takes lightly, this is a huge commitment that’s taken us years,” Rivian CEO RJ Scaringe said in an interview. “Usually, you can’t lower cost and improve performance. But here, we improved performance dramatically and simultaneously lowered cost by hundreds of dollars per vehicle.”
Two RAP1 chips will power Rivian’s next-generation on-board computer, called Autonomy Compute Module 3, which will process 5 billion pixels per second and deliver four times the performance of the Nvidia-powered system in Rivian’s current vehicles. The company’s SUVs and pickups offer driver-assistance features that require constant supervision.
Meet the Rivian Autonomy Processor.
Fast, smart, scalable and purpose-built for autonomous driving and the world of physical AI. Hitting the open road in 2026. — Rivian (@Rivian)
Automakers have for years competed to develop more capable automated-driving systems and sold investors on a future where vehicles are eventually able to offer full autonomy. However, most manufacturers have tended to count on specialist chipmakers like Nvidia, Mobileye Global Inc. or Qualcomm Inc., because developing custom AI silicon is challenging and expensive.
Nvidia — now the world’s most valuable company — dominates the market for chips in data centers used to train AI models. Its business offering automotive chips remains small, accounting for only about 1% of sales, but the company is looking to increase that share.
Tesla Inc. has been an exception to the outsourcing trend, designing its own in-car chips and making them standard hardware to help justify the investment. The Elon Musk-led company has also taken a camera-only approach, arguing that it more closely aligns with how humans drive, and that added sensors like lidar are too costly.
Rivian disagrees, taking the side of many robotaxi and car companies thatemphasize lidar’s abilityto keep tabs on a vehicle’s surroundings and back up other sensors.
“Lidar has changed a lot,” Scaringe said. “It’s no longer a big expense, it’s a very small percentage of the vehicle’s bill of materials.”
Rivian’s R2 will enter production in the first half of 2026, with deliveries beginning shortly thereafter. The initial vehicles won’t have the new chip or lidar and will therefore have more limited automated-driving capabilities.
Each of us navigates our surroundings using sight and sound together. The sensor system in the Rivian Autonomy Platform works the same way.
It takes data from cameras, radar and – coming soon – LiDAR, then fuses them together in an instant. Not as separate inputs but as one… — Rivian (@Rivian)
From 2027, Rivian will gradually introduce iterations of software that enable its vehicles to travel point-to-point without drivers needing to keep their hands on the wheel or eyes on the road.
Initially, that will be limited to highways before extending to other types of roads. Rivian’s ultimate goal is to sell customers and investors on a higher-margin software business, where personal vehicles are capable of operating autonomously without anyone in the driver’s seat.
Rivianwent publicin 2021 in one of the biggest-ever U.S. listings. It was initially hailed as a rival to Tesla after beating major automakers to market with full-size electric pickups and SUVs. Since then, however, the company hasstruggledwith the basics — its lone assembly plant in Illinois is expected to producefewer than 50,000vehicles this year, a fraction of the facility’s total capacity.
With Rivian continuing to burn cash, the company has repeatedlycut jobsand its stock is down more than 80% from highs reached shortly after the IPO. Even so, Volkswagen AG has committed almost$6 billionto a joint venture leveraging Rivian’s expertise in software and electronics, and the company continues to hire top talent from Tesla, Apple Inc. and Silicon Valley.
Hands-Free
At the core of Rivian’s current vehicles and the upcoming R2 SUV is a foundation called Large Driving Model. Its ability to learn from past and future driving will enable Rivian to upgrade the automated driving capability of older R1 vehicles — which lack lidar and use Nvidia’s Orin chips — ahead of the introduction of the more complete R2 platform expected in 2027.
The company will start charging for its Autonomy+ software platform, starting with existing R1 owners, early next year. Customers can either pay $2,500 up front to access new features over the life of the vehicle, or a $49.99 monthly subscription.
The first iteration of the platform’s new capabilities will be modest and well short of what Tesla already offers through what it markets as Full Self-Driving, or FSD.
Want more news? Listen to today's daily briefing above or go here for more info
Rivian will release an expanded version of what it calls Universal Hands Free in the coming weeks to existing owners, allowing for hands-free assisted driving across 3.5 million miles of roads, up from roughly 135,000 miles currently. But it can’t navigate or react to many traffic situations.
An updated point-to-point version of the software is expected to be ready next year, where the vehicle will be able to navigate, make turns and change lanes, while still requiring eyes on the road.
Tesla offers FSD either via a one-time $8,000 charge, or a $99-a-month subscription. The latest generation handles many everyday driving tasks, from point-to-point navigation to lane changes without hands on the wheel. The system still requires active supervision and thus is not autonomous.
