Rivian Autonomy & AI Day

RJ Scaringe - CEO Opening Remarks

Welcome everyone. We are incredibly excited to host you here today at our Palo Alto facility, which is the hub for our autonomy and technology development teams.

AI is enabling us to create technology and customer experiences at a rate that is completely different from what we’ve seen in the past. If we look forward three or four years into the future, the rate of change is an order of magnitude greater than what we’ve experienced in the last three or four years.

Directly controlling our network architecture, our software platforms, and our vehicles has created an opportunity for us to deliver amazingly rich software. But perhaps even more importantly, this is the foundation of enabling AI across our vehicles and our business.

The Evolution of Autonomy

I’d like to talk about autonomy first. The field of autonomy really started about 20 years ago, and up until the early 2020s, the approach was centered on a rules-based environment where a set of perception sensors would identify and classify objects and hand those classified and vector-associated objects to a planner that was built around a human-defined rules-based framework.

A few years ago, it became clear the approach to autonomy needed to shift. With innovations around transformer-based encoding and the design of large parameter models, the approach has moved to building a neural net-like understanding of how to drive instead of following a classical rules-based approach.

Recognizing this massive shift in how we approach autonomy, in early 2022, we began the process of a clean sheet design to our platform. The first embodiment of this work was in our Gen 2 R1 vehicles, which we launched in mid-2024.

Gen 2 Platform Capabilities

With this updated platform, our Gen 2 vehicles now have:

55 megapixels of cameras
Five radars
An inference platform that represents a 10x improvement over our Gen 1 vehicles

This platform was designed around an AI-centric approach. With the deployment of our Gen 2 R1s, we began the process of building our data flywheel to grow our large driving model.

Because this AI-centric approach represents a model trained end-to-end through the millions of miles driven on our vehicles, enhancing the perception platform or improving the compute is accretive to the capabilities of the model. The model only continues to get better as the perception and compute platform improve.

Think of it like this: if you learned to drive with bad vision and suddenly you were able to put on glasses and see much better, and then supplement that with new perception modalities like radar and lidar, and your compute—metaphorically your brain—was expanding in capability by an order of magnitude, you wouldn’t forget the things you’d learned: the rules of the road, how to operate a vehicle. But your ability to understand nuance, to respond to complex situations, and perceive the world in environments with poor or limited visibility would improve dramatically.

The Data Flywheel

Our approach to building self-driving is designed around this data flywheel, where our deployed fleet has a carefully designed data policy that allows us to identify important and interesting events that we can use to train our large model offline before distilling the model back down into the vehicle.

Gen 3 Platform for R2

While our R1 Gen 2 vehicles are using our Gen 2 sensor set and compute platform, over the last few years we’ve also been developing our Gen 3 substantially enhanced platform, and this will underpin a massive leap forward with R2.

Starting in late 2026, our Gen 3 autonomy platform will include:

65 megapixels of cameras
A robust radar array
A front-facing long-range lidar
Our first in-house Rivian Autonomy Processor

The first iteration of our in-house inference platform includes a neural engine with 800 TOPS. It’s optimized to support camera-centric AI in the physical world and enables a dramatic expansion of Rivian’s autonomy capabilities. When integrated into what we call our Gen 3 autonomy compute platform, it will deliver 1,600 TOPS.

The effectiveness and efficiency of our in-house processor has been a core focus in its development. Our Gen 3 computer is capable of processing 5 billion pixels per second.

We’ve designed this entire architecture around an AI-centric approach where the data flywheel of our deployed fleet helps the model get better through reinforcement learning. Not only does this sensor set enable a much higher ceiling than what we have in our vehicles today, it also makes the platform much better for building our model.

Feature Roadmap

We’re going to continue to see improvements on our platform. Later this month, we’ll be issuing an over-the-air update to our R1 Gen 2 customers that will dramatically expand the existing hands-free capability, going from less than 150,000 miles of roads to more than 3.5 million miles of roads in North America.

This is just a step in a series of steps:

2026: Point-to-Point Capabilities Starting in 2026, we’ll begin rolling out point-to-point capabilities in which the vehicle can drive address to address. You can get into the vehicle at your house, plug in the address to where you’re going, and the vehicle will completely drive you there.

Eyes Off The Gen 3 hardware architecture launching in 2026 expands the ceiling of what we can achieve. The next major step beyond point-to-point will be eyes off—you can navigate point-to-point with your hands off the wheel, but importantly, your eyes off the road. This gives you your time back. You can be on your phone or reading a book, no longer needing to be actively involved in the operation of the vehicle.

Personal Level 4 Following eyes off, the next major step will be personal Level 4. With this, the vehicle will operate entirely on its own. This means it can drop the kids off at school, pick you up from the airport, and be integrated into your life in ways we can’t even imagine yet today.

While our initial focus will be on personally owned vehicles, which today represent the vast majority of the miles driven in the United States, this also enables us to pursue opportunities in the rideshare space.

Rivian Unified Intelligence

Beyond self-driving, we’ve also created what we call the Rivian Unified Intelligence. This AI backbone exists across our vehicle and across our entire business. We’ve talked for a long time about software-defined vehicles, which are the foundational building block for an AI-defined vehicle where every part of the vehicle experience is designed around AI—from our Rivian assistant to enabling our direct-to-consumer sales and service model, as well as our future manufacturing infrastructure.

We could not be more excited about what we’re building, and we have a lot of details to show you here today. I’m excited to introduce Vidya to talk about our hardware platform.

Vidya Rajagopaln - Electrical Hardware Platform

Thank you. My name is Vidya Rajagopaln, and I lead the electrical hardware team here at Rivian. My team is responsible for electrical content in the vehicle ranging from our in-house 5-nanometer silicon that operates at voltages below a volt to the power electronics for electric motors that operate at 400 volts, and lots of things in between.

One common thread that runs across all these designs—beyond the fact that it involves the transport of electrons—is a Rivian ethos of vertical integration. At Rivian, we have chosen to vertically integrate critical pieces of technology that allow us to differentiate ourselves over time.

We started this journey as a startup when we consciously chose to build our ECUs in-house. Last year at Investor Day, we shared how this journey helped us get to an in-house developed zonal architecture far ahead of other OEMs.

Today, I’m here to talk about our autonomy hardware system, which is similarly very vertically integrated. As RJ shared, we will be launching our Gen 3 autonomy system late next year on the R2 vehicle platform. The hardware enabling it focuses on three main areas of leadership: sensors, compute, and overall product integration.

Sensor Suite

At Rivian, we have a multi-modal sensor strategy that provides a rich and diverse set of data for our AI models to operate on. On the R2 platform, we have:

Cameras

11 cameras providing a total of 65 megapixels of data
10 megapixels more than what we had in R1
Provides extremely rich two-dimensional data

Radar

Five radars: one front-facing imaging radar and four corner radars
Uses radio frequencies to see in total darkness while providing depth and velocity of objects
Corner radars on R2 support dual mode operation: short range and long range
In short range mode, they have very high spatial resolution, which allows us to delete the ultrasonic sensors in R2
We add sensors, but we also delete them when it makes sense

Lidar For the first time in R2, we’re adding a third sensor: the lidar. The lidar is an optical sensor, but its strength comes from the fact that unlike the camera, it has an active light source, enabling it to see much better in the dark. Another advantage is that it provides a three-dimensional view of the world, unlike cameras which provide a two-dimensional view requiring the AI models to infer depth.

In summary, camera is the main workhorse of our sensor suite, generating the bulk of the data fed to the models, but radar and lidar are critical to addressing the edge cases which would otherwise create long-tail problems.

Why Lidar Now?

There are three main factors that make this the right moment to incorporate lidar: cost, resolution, and size.

Cost: About 10 years ago, lidars used to cost tens of thousands of dollars. Today, you can get a very good lidar for several hundred dollars.

Resolution: The resolution has improved tremendously. Today’s automotive lidars have point cloud densities of about 5 million points per second, which is about 25 times better than what we could get 10 years ago.

Size: Today’s lidars are more compact and easily integrated into a vehicle, not the mechanical spinning beasts of the past.

Lidar Integration in R2

Let me show you the lidar integration. From afar, the R2 looks the same as the R2 many of you have seen and come to love. But if you zoom in closer and look up front, you can see the lidar. What you see is a seamless integration with no signs of the unsightly taxi cab-style bump or tiara structure more commonly associated with lidar integrations.

Our studio and design teams worked very closely with the supplier to shape the face of the lidar so that it blends in beautifully with the R2. And by the way, this lidar integration is camera safe—it will not burn your phone camera.

Why Build In-House Silicon?

Before we get too deep into compute, it’s important to address why we chose to build in-house silicon. It’s a non-trivial development effort. Those who’ve been involved or observed chip development efforts would know that it’s time-consuming and requires a world-class team.

The reason for doing it ties back to the same reasons for building our own in-house ECUs: velocity, performance, and cost.

Velocity: With our in-house silicon development, we’re able to start our software development almost a year ahead of what we can do with supplier silicon. We actually had software running on our in-house hardware prototyping platform well ahead of getting first silicon. Our hardware and software teams are co-located and able to develop at a rapid pace that’s simply not possible with supplier silicon. This means we’re able to get to market sooner with the most cutting-edge AI product.

Performance: We understand our application and our vehicle architecture thoroughly and are able to optimize our silicon for our use case. We don’t just design for today’s use case—we design with headroom for the models of the future. By building purpose-built silicon, we don’t carry the overhead from leveraging a design built for some other task and repurposed for autonomous driving. We built this silicon so it would do a really good job at autonomous driving and physical AI problems. This enables us to get the best performance per dollar spent.

Cost: When we design in-house, we’re able to get the best cost point and power point. The cost reductions come from the fact that this is optimized for our use case—not just the chip use case, but the whole vehicle use case—as well as the meaningful reduction in supplier margins.

Gen 3 Autonomy Computer

Our Gen 3 autonomy computer is the next step in our vertical integration journey and features our very own Rivian-designed custom silicon. It is a highly integrated solution. There is very little on the board beyond the two instances of Rivian silicon, power supply, and passives. The hardware and software on this computer are fully designed and developed by Rivian.

This computer achieves four times the peak performance of our Gen 2 computer while improving power efficiency by a factor of 2.5.

RAP One - Rivian Autonomy Processor

The Rivian Autonomy Processor, or RAP One as we call it, is the first in a series of silicon built for physical AI. It’s actually much more than one piece of silicon—it’s a multi-chip module (MCM) that integrates Rivian silicon and memory die.

Technology: Our custom Rivian silicon is produced on a leading-edge TSMC 5-nanometer automotive process.

Neural Engine: The star of the die is a Rivian-designed neural engine capable of 800 sparse int8 TOPS (trillion operations per second).

Scalability: The chip was designed with the intent of providing different cost and performance points. We can put multiple RAP processors together in a system, and they can talk to each other via a custom high-speed link we call Rib Link.

Multi-Chip Module Design

Let’s take the lid off RAP One. What you see is the RAP One SOC in the middle surrounded by three memory die spread across two sides. This allows for three independent LPDDR5 channels but, more importantly, allows for very tight integration between the SOC and memory, enabling a very clean data eye which enables high memory bandwidth.

With RAP One, we’re one of the first to introduce multi-chip module packaging for high-compute applications in automotive. This is not to be confused with systems in packages (SIPs), which have existed in automotive for a very long time.

Memory bandwidth is key for AI applications, and this tight coupling enables us to achieve a net bandwidth of 205 gigabytes per second. The MCM design also enables us to significantly simplify the design of the PCB, resulting in meaningful cost reduction.

SOC Architecture

The SOC itself is designed to solve the needs of autonomous driving:

Neural Engine: Rivian-designed, capable of 800 int8 TOPS

Application Processor: 14 power-efficient ARM Cortex A720 AE cores, allowing us to leverage the rich open-source software ecosystem. We will be the first OEM to introduce the ARM V9 compute platform for automotive using the Cortex A720 AE in production vehicles.

Safety and Real-Time Processing: High-availability safety island and real-time processing unit built using eight ARM Cortex R52 cores

Sensor Processing: Image signal processor, encoder, GPUs, etc.

Neural Engine Capabilities

The Rivian neural engine is designed to implement state-of-the-art deep learning models for perception, control, and planning. It is flexible and supports mixed precision data formats with native hardware support for:

Transformers
All types of attention (multi-headed attention, deformable attention, and more)
Non-linear functions like softmax
CNNs
Lidar and radar processing (which can be very unstructured)
Weight decompression to relieve pressure on memory bandwidth
Concurrent execution of up to four models at any given time

We made a significant investment in the development of tools and a middleware stack to exploit the hardware. The entire software stack is fully developed in-house, including:

In-house compiler that can take standard models and generate code targeting our neural engine
Profiling tools to help users optimize their code
In-house middleware stack that is target-agnostic (same middleware on Gen 2 and Gen 3 platforms)

Functional Safety

What makes silicon for physical AI different from general silicon targeting inference is the importance of functional safety. RAP One was designed from the get-go to factor in functional safety in every block of the design.

We adhere to the ISO 26262 scheme for risk classification (ASIL - Automotive Safety Integrity Levels). Every block is designed at the appropriate ASIL level, and hardware and software are implemented to ensure that level is achieved. Even our chip-to-chip interconnect, the Rib Link, is protected using this scheme.

In some cases, this means putting extra redundant hardware in the chip that does the same function twice and cross-checks the results. In other cases, it means putting ECC on memories instead of parity. We also have software that runs on the chip when it’s working in the vehicle and at key-on to ensure the chip is still functionally safe.

Scalability and Flexibility

The RAP One chip is designed to be scalable. While the first instantiation is a two-chip solution targeting autonomy in the R2 vehicle platform, it can be easily extended to solve other physical AI problems such as in robotics. It can scale down to a single-chip solution for low cost or scale up to multiple chips for more performance.

Rib Link was specifically designed to allow multiple RAP chips to talk to each other via a high-bandwidth, low-latency interface at data rates of up to 128 gigabits per second, allowing sensor data from one SOC to be seamlessly shared with other SOCs.

The scalability doesn’t end there. RAP One was designed to also be flexible in configuration. While the system to be deployed in R2 is liquid-cooled, we have demonstrated that it can be configured as an air-cooled system.

Performance Validation

I’m happy to share that we have successfully demonstrated that our silicon is robust and meeting the performance goals we set out at the start of the project. While peak TOPS are useful to indicate hardware capability, a more useful measure is the ability of the system to process sensor data.

We have shown that a Gen 3 autonomy hardware system is capable of processing 5 billion pixels per second of sensor data.

We are very proud to be at the leading edge of multi-modality sensing and to be continuing our trajectory of vertical integration with our RAP One chip and Gen 3 autonomy computer. We expect that at launch in late 2026, this will be the most powerful combination of sensors and inference compute in consumer vehicles in North America.

We are now actively testing the silicon, systems, and vehicles. For those attending in person, you’ll get a chance to see some of our subsystem test boxes that test the entire hardware-software configuration. We have also integrated the hardware into our R2 vehicles and are continuing to test it extensively.

I will now hand it over to James, who will show you how his team is continuing to improve the autonomy experience for our customers and how he plans to harness the power of RAP One to make autonomy better.

James - Autonomy Software Platform

Thanks, Vidya. Vidya just discussed all the amazing autonomy sensors and compute we’ll have on R2. Now I’ll go into detail on some of the software that runs on them and powers the Rivian autonomy platform.

Large Driving Model

Our large driving model is trained end-to-end from millions of miles of driving sequences collected across our fleet—directly from pixels, radar returns, and lidar points to trajectories.

This large driving model uses state-of-the-art techniques based on:

Transformers
Auto-regressive prediction
Reinforcement learning

This turbocharges our velocity by allowing us to leverage innovation from the world of large language models. It’s also built entirely in-house, which gives us unprecedented flexibility in being able to change all parts of the stack. We don’t need to coordinate with other tier ones and tier twos to make changes.

Consequently, our features improve with every update. Finally, and most importantly, the autonomy platform is built on a data flywheel where growth in vehicle fleet and feature adoption drives improvements in autonomy that compound over time.

Multi-Modal Sensing and Early Fusion

Let’s look under the hood and discuss the data flywheel in more detail. We start with a multi-modal onboard model that runs on every customer’s vehicle. The goal for our onboard sensing stack isn’t just human-level—it’s superhuman-level. Multiple modalities enable that, allowing our vehicle to see way beyond what a person can.

By end-to-end training, the sensor data is early-fused into a singular world model. This is a system where the sensors complement each other—they don’t fight against each other. Just like being able to hear can make you a better driver, multiple sensors can make Rivian autonomy better, enabling enhanced precision and more confident predictions.

With more sensors comes richer and better fidelity data. More sensing modalities allows us to achieve the same level of accuracy as a uni-modal system but with much less data, or to surpass the uni-modal system with the same amount of data. It’s a very efficient approach.

Visualizing Early Fusion

Let me show you how this works in practice by visualizing the output of early fusion.

Cameras Alone: Our cameras are really good—some of the highest combined megapixel counts and dynamic ranges of any vehicle on sale today. When cameras clearly sense things, the system works very well. In fact, cameras alone can handle most autonomy tasks most of the time. But for full autonomy, most of the time isn’t enough. Autonomy needs to work all of the time—on a moonless night, in the snow, and in fog. In those cases, cameras alone don’t cut it. If we can’t sense something, we can’t expect the system to handle it.

Adding Radar: This is where we are with every Gen 2 R1 vehicle today. You can see that we’re now able to detect much more. The system can detect more occluded objects, can assign better depths to objects, and is better at estimating object velocities. The confidences and redundancies are also much greater than with a camera-only system.

Adding Lidar: To unlock full autonomy, we need to go further. Adding lidar creates the ultimate sensing combination. It gives the most comprehensive 3D model of the space the vehicle is traveling through. The combination of all three sensors identifies more objects and can detect things more quickly. This trinity of modalities enables autonomy features like eyes-off driving and personal L4 by increasing perception quality, safety, and redundancy.

Network Architecture

Our onboard network is designed from the ground up to flexibly incorporate new modalities and new sensors. Together with our in-house silicon team, we’re co-designing this network to run optimally on our chip.

Each pixel, each radar return, each lidar return is encoded, projected, and then combined into a geometric feature space. This is where the fusion happens—optimally and automatically learned through end-to-end training. There’s no extra complexity added. There’s no handwritten rules that need to adjudicate. The network has figured out the best way to combine this information.

World Model Output

This whole fused tensor is fed into our transformer-based decoders to produce the world model. From the same input, the network is trained to generate all of these different outputs for the world around the vehicle:

Other Objects and Agents: Detection and tracking of vehicles, pedestrians, cyclists, etc.

Dense 3D Occupancy: To handle short-range maneuvers and narrow negotiation

Local Map: Note that this is quite different from typical robo-taxi efforts where the map is pre-generated offline and then localized online. Those maps are expensive to generate and hard to maintain. But just like a human driver can navigate a road they’ve never seen before, our local map is produced and accumulated directly from what the vehicle perceives.

Estimated Trajectories: These represent the model’s best estimate of how to proceed through the scene. This output will be a key technology enabler for point-to-point driving.

Autonomy Data Recorder (ADR)

Now we’re moving to the next stop in our flywheel: the Autonomy Data Recorder. This is the system that turns real-world driving into data. Essentially, our entire Gen 2 fleet becomes a huge queryable dynamic database of driving scenarios.

The data recorder runs trigger code that can fire off any set of signals seen by the world model:

Jaywalking pedestrians
Red light runners
Large animals in the road

These are all examples of cases which you can mine with triggers.

We can also run more general triggers, such as finding divergences between the human-driven trajectory and our predicted trajectories. These might indicate areas where the model could be improved.

We can even push new triggers live to our fleet outside of the usual OTA cycle. This allows us to capture the data we need on demand with minimal turnaround time, greatly speeding up development. That’s a huge driver of developer productivity. As soon as an engineer wants to find more scenarios of an event, they can mine for them immediately.

Because ADR is so selective, it’s also very efficient. The vast majority of boring driving data is never captured, never uploaded, and never trained on.

Once a trigger fires, all the sensor data before and after the event is captured, tagged, compressed, and uploaded. That data is then immediately available to engineers. These scenarios can be used for model training, evaluation, or replay through our simulator.

Here you can see examples of scenario clusters found automatically by our ADR system:

Environmental conditions such as dusk and night
Map-based scenarios such as tight turns and merges
Agent-based situations like animals, bike racks, and semi-trucks

After upload, all of these curated scenarios are immediately available on the Rivian cloud. As the fleet expands and adoption increases, the size of our knowledge base is growing and compounding. From next year, this growth will be further accelerated with the additional volume of R2.

All of this data is stored securely. We don’t associate any sequences with your VIN or, if you select it, your home or place of work. Through ADR, every sequence is already tagged without further processing.

Because we store all sensor modalities, the data is incredibly rich and complete. This allows us to auto-label most sequences using large offline models, which would be too slow to run onboard. In fact, the vast majority of our training data today is auto-labeled. That’s massively more efficient than using human annotators.

Ground Truth Fleet

You may have seen other autonomy providers’ ground truth vehicles. They typically have multiple non-production sensors such as lidars strapped to the roofs and sides of the vehicles. These fleets are incredibly valuable for training perception systems. The lidar data is so crisp, it’s essentially used as ground truth to train the other production sensor sets. But because they’re prototypes, these fleets are typically small, numbering in the tens to hundreds.

In contrast, for the lidar-equipped R2, every vehicle will become a ground-truthing vehicle. That’s orders of magnitude more data than other OEMs—an incredible force multiplier for better and richer training data that massively accelerates our progress.

Large Driving Model Training

Let’s see how we use all of this data to benefit our customers. This happens in the Large Driving Model (LDM). The LDM is an end-to-end model from sensors to driving actions, and it’s based on many of the same technologies used in large language models.

Transformers: LDM uses neural net transformers for processing, just like with LLMs.

Tokens: The large driving model uses tokens for training, also just like LLMs. But instead of thinking about these tokens as words, they’re actually small parts of trajectories that are jointly predicted and assembled together.

Reinforcement Learning: The large driving model also uses reinforcement learning, just like state-of-the-art large language models. But here, instead of aligning the output with human values and intentions, we align the LDM output for safe, performant, and smooth driving.

Because LDM is such a close cousin to an LLM, we can reap all of the advancements, investments, and innovations being made in improving generative AI and apply them directly to our driving task. This makes the LDM incredibly cost-efficient to develop.

Reinforcement Learning Deep Dive

Let me do a deep dive into how LDM is trained by reinforcement learning. Here’s sensor data from a scenario where we’re approaching a stop sign. That sensor data is fed into our transformer-based encoders. Then we sample multiple trajectories from this model, token by token and trajectory by trajectory. Different tokens shape the trajectories in different ways.

Once we’ve sampled all these trajectories, we then need to rank them. One trajectory might be the most human-like—the vehicle slowed almost to a stop but then rolled it. A lot of our Rivian data is like that. Another might be stopping too soon. And one is just right—stopping behind the line and correctly following the road rules.

What we’re able to do is apply our road rule rankers that can then say the optimal trajectory is the correct one, and then we reorder them. Through backpropagation, the model is trained to produce more of these types of trajectories in similar scenarios in the future.

That’s highly simplified, and you’re just looking at one scenario here. But imagine this process running millions of times a second across millions of scenarios with a whole database of road rule costs and losses. That’s how LDM is trained. We can then distill this model into one that we can run onboard.

Validation and Release Process

All of this work results in new models, continuous enhancements, refinements, and new features that we continuously deliver to our customers. But how do we know we can release?

Cloud-Based Simulator: We’ve built a cloud-based simulator that runs the whole autonomy stack through millions of miles of real-world scenarios on every release. This allows us to measure safety, comfort, and performance in a statistically significant way without having to manually drive those miles.

Apprentice Mode: We also have a capability we call apprentice mode. Before we release features, we can launch them in the background of a previous release. We can then monitor the performance of that new version compared to the human-driven miles, but also compared to the previous version of autonomy. That allows us to do an even bigger evaluation in the tens of millions of miles.

Through simulation and apprentice mode, we can rapidly build the confidence we need to ship new features and enhancements to customers.

Because the system is developed entirely in-house, we can update any part of the stack from the lowest-level camera drivers all the way to the highest-level motion planning code. That means the whole stack is always improving with every release, and we have a feature roadmap that stretches to the highest levels of autonomy.

Universal Hands-Free

When we surveyed customers this year on the autonomy capabilities they wanted most, the answer was resounding: they wanted more road coverage for hands-free highway assist. Previously, we supported 135,000 miles of divided highways. But as RJ mentioned, our map is about to grow.

Universal Hands-Free unlocks over 3.5 million miles of hands-free driving on roads across the US and Canada. If there’s a painted line and it’s clearly marked, you can now drive hands-free.

Autonomy Plus

Universal Hands-Free will be part of our paid tier bundled into one simple package: Autonomy Plus. It’s a one-time fee, or you can pay month-to-month. Autonomy Plus features will be available to all Gen 2 customers for free until March next year.

This is just the beginning for Autonomy Plus. We have many exciting features on the way, such as:

Point-to-point
Automatic parking
Eyes-off (enabled by the lidar on R2)

As our fleet continues to grow and adoption continues to increase, our data flywheel will continue to grow. We’ve been thinking about this as a circle, but in fact, the system is better on every orbit. A better analogy is an upward helix, continually improving and compounding on itself.

With that, I’d like to hand over to Wassym to discuss some of the other improvements being made in AI here at Rivian.

Wassym - Rivian Unified Intelligence

Thank you, James. We have made significant progress in our AI-enabled autonomy stack. But as James said, it doesn’t stop here. AI runs through the core of everything we do. It’s a profound platform shift which changes our product and everything we do at the company—from the way we design, we develop, we manufacture, and we service our cars.

This is all made possible by the Rivian Unified Intelligence, a common AI foundation that understands our products and operations as one continuous system and personalizes the experience for our customers.

How It Works

AI-Ready Vehicle OS: We revamped our vehicle operating system to be AI-ready.

Multi-Agent Platform: We developed an in-house multi-agent, multi-LLM, multimodal intelligence platform. The platform is built on a robust data governance framework with security and privacy as main tenets.

Specialized Agents: We have a suite of specialized agents. Every Rivian system—from manufacturing, diagnostics, EV trip planning, navigation—becomes an intelligent node through MCP. The beauty here is we can integrate third-party agents, completely redefining how apps in the future will integrate in our cars.

Model Orchestration: We orchestrate multiple foundation models in real-time, choosing the right model for each task.

Memory and Context: We support memory and context, allowing us to offer advanced levels of personalized experience.

Native Multimodality: The architecture is natively multimodal, using audio, vision, and text through the same unified layer.

Edge AI: The beauty of our architecture is the seamless integration between the cloud and the edge. Edge AI with an embedded small language model allows us to achieve higher levels of performance, lower latency, and the best conversational experience.

R2 will have close to 100 TOPS of AI fully dedicated to in-cabin experience. This will allow us to move most of the intelligence workloads from the cloud to the edge, powering an in-cabin AI experience fully available when the car is offline.

Applications Across the Business

The Rivian Unified Intelligence is the connective tissue that runs through the very heart of Rivian’s digital ecosystem. This platform enables targeted agentic solutions that drive value across our entire operation and vehicle lifecycle.

Factory - Diagnostics Agent: Our diagnostics agent is the ultimate example of unified intelligence in action. It instantly connects real-time telemetry from vehicles on the assembly line and allows us to validate quality at production, identifying changes needed before the vehicle leaves the factory.

Service: The same unified intelligence is fundamentally helping us redefine service. Because the platform is grounded in real vehicle data, every Rivian technician is now being empowered by AI systems trained on live software data, service manuals, repair histories, electrical diagrams, and supply chain logistics. This is accelerating service repair time by hours, helping us dramatically improve technician efficiency.

It doesn’t stop with our technicians. The same platform will enable, in the future, our customers to self-troubleshoot and resolve minor issues directly from the Rivian mobile app and from your Rivian car.

The Rivian Assistant

Now let me tell you more about the in-vehicle experience. This architecture will fundamentally reshape how we all interact with our Rivians.

Today I am very excited to announce the Rivian Assistant.

The Rivian assistant is fully integrated into the user experience and our in-car operating system. It’s designed to understand you, your vehicle, and the context you’re in.

The Rivian Assistant will be available for all Rivian Gen 2 and Gen 1 customers in early 2026.

Live Demo

Rather than tell you about all its features, we’ll do a live demo from the R1S Borealis. Oscar will be our main driver today. Oscar is the product manager behind the assistant.

[Demo begins with Oscar in the vehicle]

Oscar: To initiate the Rivian assistant, you can either hold the left steering wheel button or just say, “Hey, Rivian.”

The assistant is not just an integration of a chatbot on top of the vehicle UI. We built an agentic framework allowing us to integrate into the larger ecosystem and bring your own digital context to the car.

We started with Google Calendar. We had a ton of fun collaborating with the Google team on this project.

Oscar: “What’s on my calendar today?”

Assistant: “You have two events today. Call with Tim from 2:00 p.m. to 2:30 p.m. and meet up with Wassym from 3:00 p.m. to 4:00 p.m.”

The agentic integration allows us to not only connect with your calendar and read it, but also take actions and manage it.

Oscar: “Can you move my call with Tim to 5:00 p.m.?”

Assistant: “I’ve moved your call with Tim to 5 p.m.”

As you see, the same action would have probably taken me multiple tabs, clicks, and swipes to get to the same result. The assistant can help you control your calendar in a much safer and easier way.

Google Calendar is just the beginning. The platform will expand to many more applications and third-party agents as they become available.

Beyond the agentic integration, our AI platform can blend your personal context into your vehicle context. In this case, the Google Calendar agent is connected to other vehicle applications.

Oscar: “Let’s go to my meet up with Wassym.”

Assistant: “Navigating to Ferry Building. You’ll drive for 55 minutes and arrive around 10:50 a.m.”

The AI integration goes much deeper than navigation. Because this is an AI-native operating system, it can connect with all vehicle apps, including our customers’ favorites like EV trip planner or drive modes.

Oscar: “How much battery will I have when I get to my destination?”

Assistant: “You will have 67% battery remaining, which is about 231 miles of range.”

Oscar: “Can you switch to a more efficient drive mode?”

Assistant: [Switches to Conserve mode]

As you see, Oscar did not have to specify “Conserve mode.” This is the future of vehicle UIs. The assistant allows users to perform use cases without knowing the exact vehicle command, without knowing where they are in the different menus in the UI, making it way easier for every user to interact with the Rivian.

Oscar: “Can you make the seats toasty for everyone except me?”

Assistant: [Turns on seat heating for all seats except driver]

Imagine the possibilities that this opens up. The assistant takes the vehicle experience to the next level. Instead of having multiple UI commands, multiple taps on the touch screen, you can perform the whole task with just one natural language command.

Messaging Integration

One of the features that our community has requested is messaging. I really wanted to get messaging right.

Oscar: [Receives a text message]

Assistant: “Read my last message from Wassym.”

Message: “Hey, I’m nearby. Can you find a good restaurant near the ferry building? Let me know.”

Assistant: “Do you want to reply to Wassym?”

We didn’t only want to stop at reading your message. The messaging app with this new AI platform is fully integrated into the vehicle operating system and has access to every single application and control.

Oscar: “Can you find some restaurants near my destination?”

Assistant: “I found Terrene about 30 miles away in San Francisco. It has a 4.2-star rating. Would you like to go there or explore other options?”

Oscar: “Actually, can you send a text message to Wassym and show him the top three options from this list? Ask him which one he wants to meet up at and include my ETA.”

Assistant: “Would you like to send a message: ‘Hey Wassym, I found these restaurants near Ferry Building: 1. Terrene, 2. Eclipse Kitchen Bar, 3. Super Duper Burgers. Which would you prefer? I’ll arrive around 10:53 a.m.’”

Oscar: “Yeah, send it.”

Assistant: [Sends message]

You just saw the difference that native integration makes. The assistant has memory, has context—it remembers the full story: who you’re talking to, where you’re going, and what you just searched for. And then it puts everything into a perfect message.

The Future of AI-Defined Vehicles

What you’ve witnessed today is more than just a new voice assistant. This is a peek into the Rivian Unified Intelligence platform that powers the new foundation of Rivian’s digital ecosystem.

The gap between software-defined vehicles and traditional architectures is getting exponentially wider with AI. Rivian is uniquely positioned to move from a software-defined vehicle and bring to the world an AI-defined vehicle.

RJ Scaringe - Closing Remarks

That was so much fun to watch. So much time and effort have gone into building the platforms, designing the architectures, building the teams, growing the teams, organizing the teams—all the work that goes into these really complex systems.

I often describe it as if you’re building the plumbing. If you’re building a house, you don’t start with the finished house. It takes years of planning. You have to do foundation work. You have to do wiring and plumbing that go into the house. And then at the very end, it all comes together.

The work we talked about on our in-house processor—this is something that has been years in the making. The amount of effort and time that’s gone into it, and by the way, the amount of effort that went into this not leaking, which is amazing, is just so inspiring.

I spent some time last night with the team talking about this right before we’re about to show it today. One of the lead engineers looked at me and said, “Boy, we’ve been working on this for years, and I haven’t been able to talk about it. It’s so cool. Tomorrow I can start to talk about what I do every day, all day long.”

Between the work that we put into the processor, the large driving model, how that feeds our data flywheel, this large flywheel approach to building a dataset that continues to improve our model, and all the work that went into first building a software-defined architecture, developing and building all the electronics that go into the vehicle, and then using that as the foundation for enabling an AI-defined vehicle—this is coming together today. You’re seeing the house start to form in front of you.

What customers are going to see on our Gen 2 R1 vehicles starting very soon is a lot of these features. As I said, later this month we’re going to be growing the amount of miles you can access Universal Hands-Free from just under 150,000 to 3.5 million miles.

2026: Point-to-point navigation

Shortly thereafter: Hands off, eyes off

We’re very excited, and we appreciate all of you being here today.

For those that are here, we have demos. I know there’s a lot of people here, so not everybody will get to do a demo, but for those that are doing demos, you’ll get to see the point-to-point navigation and the work that’s gone into that. We have a bunch of great displays that show some of the hardware.

For those that aren’t here, we appreciate you listening along and we appreciate your support and enthusiasm for what we’re building.

Thanks, everyone.