The Brains of Machines That Move: 3 Edge AI Stocks Built for Robots and Cars

The claim that intelligence is leaving the data center shows up most dramatically in machines that actually move — robots and cars. A tenth of a second of delay can turn into a crash, so there is no time to ask the cloud. The decision has to finish inside the machine.

Here are the three names I treat as core to this category. They own, in turn, the brain of the robot, the brain of the car, and the eyes of the car.

1. Nvidia — the brain that rides inside the robot

Plenty of people know Nvidia only as the data-center company. What I watch instead is the chip it is pushing into machines that move.

Jetson Thor is an on-robot brain small enough to ride inside a humanoid, carrying about seven times the compute of the chip it replaced. On top of it sits Isaac software and a family of robot foundation models called Groot that teach a machine how to see and how to move. The names already building on it read like the whole field: Amazon Robotics, Boston Dynamics, Figure, and Caterpillar.

The numbers explain why you cannot ignore this company. Nvidia's annual profit has multiplied more than 40 times in six years. Six years ago it earned under $3 billion in a year; in its most recent year it earned $120 billion, at a gross margin above 70%. I consider that one of the greatest runs in the history of business. And robotics is the next engine being bolted on while that machine keeps compounding.

2. Renesas — a supercomputer inside the car

Renesas is the world's largest maker of the microcontrollers that quietly run everything inside your car, from the engine to the dashboard. Now it is putting real on-device AI into that same silicon.

The flagship is the R-Car X5H — the first automotive chip ever built on a 3-nanometer process. It delivers up to 400 trillion AI operations per second (400 TOPS) while drawing roughly 35% less power than the chip before it. That is effectively a supercomputer for a car. Bosch and Zephyr, two giants of the auto-parts world, are already designing on top of it.

The underlying business is genuinely solid. Operating margins run in the high teens, so this company actually makes money, and it bought design-software maker Altium and edge-AI startup Pictorus to own the whole workflow, from the tool a customer designs with down to the chip it runs on. The most recent headline showed a loss, but that was a one-time write-down on a silicon-carbide investment, not the business itself. One heads-up: in the US it only trades over the counter, which makes it a little harder to buy.

3. Mobileye — eyes already inside 230 million vehicles

Mobileye is the purest way to own automotive edge vision. While everyone else bolts AI onto the dashboard, Mobileye builds the EyeQ chip that is the eyes and reflexes of the car, processing every camera frame inside the vehicle itself.

The scale is staggering. EyeQ is already built into more than 230 million vehicles on the road — a number that quietly makes Mobileye one of the most widely deployed pieces of AI hardware on the planet.

This one is also a turn. A giant one-time write-down buried the numbers and is only now clearing out, while revenue climbs again, up roughly 27% in the most recent quarter as EyeQ volume comes roaring back. Mobileye has also stopped being only the chip supplier; it is standing up its own robotaxi fleet, owning and operating the cars itself. The one risk I will not gloss over is structural: Intel still controls a large block of these shares and can sell them into the market at any time. But that 230-million install base is a moat rivals cannot conjure overnight.

How I frame the three

The three have different personalities. Nvidia is a dominant profit engine with robotics as an option on top; Renesas is a steadily profitable automotive champion; Mobileye is a turnaround shaking off a write-down as volume returns.

From my seat, the common thread is one thing: in domains where latency is risk, the law of physics that forces the decision to finish on-device is what builds these moats. If the data center was act one of the edge AI shift, machines that move are the front line of act two, where that intelligence is needed first.