Rivian founder and CEO R.J. Scaringe hosted the event EV The manufacturer's first Autonomy and Artificial Intelligence Day this week, announcing a slew of major achievements for its no-longer-young company. At Rivian's Silicon Valley headquarters, the automaker revealed a project it has been keeping secret: a silicon chip of its own design.
The chip is the processor used in the next version of Rivian's on-board computer. Dubbed Autonomy Compute Module 3, it is capable of 1600 sparse INT8 (8-bit integer) TOPS (trillion operations per second) and 5 billion pixels per second of processing power. Without going too deep into bits and bytes, these numbers indicate strong performance.
Rivian talks about data with numbers that boggle the mind. In terms of scale, Rivian says this new setup will quadruple the capabilities of the Nvidia chip-based system it currently uses.
Rivian is focusing on a neural engine and a new middleware stack.
Semiconductors are the brain that now controls almost all the digital equipment in our lives, from smartphones to cars. Making chips typically requires a multibillion-dollar cleanroom facility and an incredibly complex process that produces tiny wafers of silicon. That's not what Rivian is doing; The automaker buys the chip itself, but the design and housing are made in-house by Rivian. A couple of years ago, developing your own chip was just a dream, but this is a huge advantage.
“We are conscious of the fact that we are a car company, not a company that constantly produces chips,” says Vidya Rajagopalan, senior vice president of electrical equipment at Rivian. Rajagopalan worked on Model 3 at Tesla and several chip and systems companies before joining Rivian in 2020, and she knows what she's talking about. Rivian runs on ARM and uses the company's microprocessor, and Rivian developed the core, the neural engine. This is the most important part of the chip, and where Rivian brings the most value, Rajagopalan said.
“Creating a chip takes a long time and requires a world-class team,” says Rajagopalan. “But the benefits are speed, performance and cost. This means we can get to market faster with cutting-edge AI product and we will be able to optimize our silicon for our use cases, leaving room for future models. We don’t incur the overhead of a design intended for a different purpose.”
In other words, designing the chip allows Rivian to tweak the system as it goes along, rather than getting a generic chip and figuring out how to fit it. Customizing the use of artificial intelligence is a core tenet of the company's roadmap, underpinning its software, research and autonomy mapping, and Rivian Assistant, its new voice command setup. Wake it up with “Hey Rivian,” and the system can handle complex, multi-part requests, interrupts, and a text-based interface that eliminates the need for Apple CarPlay and Android Auto.
Another aspect of the equation is Rivian's new middleware stack, also developed in-house. Middleware is the glue that binds the pieces together, acting as a bridge to connect different applications, databases and services. It maximizes flexibility and speeds testing and development by scaling across a variety of platforms and computing hardware.
Rivian Moves Forward with Plan to Use AI Everywhere
Rivian also unveiled its next-generation standalone platform, which will run on new chips. The proprietary, purpose-built silicon has been developed to “achieve significant advances in self-driving technology,” Scaringe says, as part of his roadmap to transform the future of the industry with artificial intelligence.
“AI allows us to create technology and customer experiences at a speed that is completely different from what we have seen in the past,” Scaringe says. “If we look three or four years into the future, the rate of change will be an order of magnitude greater than the entire experience of the last three or four years.”
As the market debates a potential AI bubble that could collapse like the dot-com bubble of the late 1990s, technologists, policymakers and environmentalists are voicing their concerns. AI, despite all its potential, also poses a threat to the environment due to huge energy demands and job losses.
“Integrating and implementing AI in real-world settings can be challenging and lead to undesirable results as we move forward,” says Ali Shojaei, a professor at Virginia Tech. “For example, the impact of AI on the environment and energy consumption cannot be overlooked. Privacy and data security are also valid concerns as AI becomes more widely used and sensitive information is automated.”
Scaringe insists we are at a technological inflection point.
“The way we approach artificial intelligence in the physical world has changed dramatically, and the idea of not having full-blown artificial intelligence in every aspect of our lives will be almost unimaginable,” the CEO predicted. in a video released this week..
About five years ago, Scaringe said, the approach was based on a rule-based environment with a set of perceptual sensors to identify and classify objects. A few years ago, it became clear that we needed to move to a neural network understanding of how to drive a car.
All this will be implemented in the upcoming R2 model with Rivian Autonomy Processor 1 chips and a new LiDAR sensor combined with cameras and radar technology. For example, Waymo's self-driving cars use LiDAR sensors around the entire perimeter of the car, sending laser pulses in all directions to detect objects. Rivian's main lidar sensor is built into the car above the windshield instead of a Waymo-style dome that screams “taxi.”
Scaringe's updated vision for self-driving Rivians kicks off in 2026, when the automaker brings point-to-point navigation to the R2 and through over-the-air updates to its second-generation vehicles. Rivian is clearly aiming for a self-driving experience that doesn't require the driver to pay close attention to the road without having to participate in the operation of the vehicle. And after that, according to the CEO, there will be Level 4 autonomy, meaning the car can take the kids to swim practice for you.
Rivian engineers acknowledge that autonomy is a work in progress, and each R2 vehicle will be eligible for crowdsourced training on the system using artificial intelligence. When asked about numerous cases Waymo cars illegally pass school busesDirector of Autonomy Products and Programs Nick Nguyen noted that the driver is still responsible in Level 2 autonomy situations. This is not level four yet.
“We won't be able to handle every situation a car might encounter, but if a person is looking at the road [which is required at this level]in such a situation, the driver must take control,” Nguyen emphasizes.
The company will begin charging for its Autonomy+ software platform next year; customers can either pay $2,500 up front or a monthly subscription of $49.99. That's less than Tesla's FSD system, which requires either $8,000 one-time or $99 per month.
PopSci Gift Guide for 2025
