Opinion ยท By Jan Cortenbach
There are 45 million tiny chips out in the world right now, running AI on doorbells, dashboards, drones, and factory cameras. Most people who own one don't know it. Most investors who own the stock don't fully know what they've bought. It's called Ambarella. And it just quietly became the largest independent edge AI chip company that isn't run by NVIDIA or a hyperscaler.
Here's the pitch. The chip that already sees everywhere doesn't need to convince anyone. Ambarella already won its market. Now that market is expanding into every physical AI application on the planet. That's the thesis, and here's why it holds up.
What edge AI actually means
Cloud AI is what runs ChatGPT, the massive data centers doing the heavy lifting far away. Edge AI is when the chip in the device does the thinking itself. A security camera that recognizes a person without sending video to a server. A car that spots a pedestrian in 50 milliseconds instead of waiting for the internet. A robot that sees a broken pallet and adapts. Edge AI is faster, cheaper, and more private. It's like the difference between calling a friend for help and already knowing the answer yourself.
Ambarella owns this space
FY26 revenue came in at $390.7 million, up 37% year over year. Roughly 80% of that revenue comes from edge AI applications. Their cumulative edge AI revenue just crossed $1 billion. They have 12 different AI chips in production, supporting more than 370 unique customer projects, running over 200 different AI model architectures. That's not a science project. That's a platform business with switching costs.
Then there's the Hanwha deal
Announced in Q1 FY27, it is an agreement worth more than $800 million over more than 10 years to jointly develop edge AI silicon for security cameras, robotics, industrial automation, and life sciences. That's the kind of deal that quietly takes the risk out of a whole decade of revenue. Hanwha is one of the largest security camera makers in the world. Locking them in for 10 years is not something you do with a shaky product.
The physical AI angle nobody's talking about
The same chip family that runs the camera on your Ring doorbell is what a humanoid robot needs to see the world. Ambarella already has the manufacturing scale, the software stack, the customer relationships, and the low power silicon that a battery powered robot desperately needs. Building an AI accelerator for humanoids from scratch is like building a car company from scratch. Possible, but why would you when the parts are already there?
The risk sheet is honest
Consumer IoT, meaning dashcams and action cameras, is declining as that market matures. Gross margins compressed a bit last quarter. And the stock trades at a premium, with analyst estimates putting the forward P/E around 50x. That leaves no room to disappoint. If AI adoption at the edge slows, the multiple contracts fast.
But look at what's happening structurally
Automotive revenue is at record highs, driven by AI in commercial vehicles. The chip design cycle is spinning off higher margin AI SoCs that command better prices. And robotics is a stated growth vertical that management is investing behind. This isn't a company hoping for the AI wave. It's a company that already caught it and is now widening the ride.
Compare it to Ouster or Aeva, sensor plays that still burn cash. Ambarella is already profitable, already scaled, and already has the customers. It's not a bet on physical AI happening. It's a bet on physical AI happening on top of hardware they already ship.
The high timeframe picture
The tape tells the same story the fundamentals do. After a long base in the fifties and a washout to the high forties in 2025, the stock has broken out toward the mid nineties, near the top of its trailing year range. The market is starting to price in the edge AI and robotics thesis, which is exactly why the risk sheet above matters. You are no longer buying this quietly.
The most valuable chip company in your basket might be the one you've never heard of. That's usually how it works. By the time it's obvious, the trade is over.
Opinion by Jan Cortenbach. Not investment advice. Verify all numbers against primary filings.