New Faces in the Embedded Classroom — What Two Weeks of Teaching ARM Showed Me

To Be Honest, I Didn't Expect Much

I run a small embedded development company. BMS, automotive software, devices running on Linux and RTOS — building software that touches hardware is what I do. In an earlier post, I described this market as one that schools rarely teach and people are reluctant to enter.

That wasn't an exaggeration. Until recently, interest in embedded software was scarce — in the market and among students alike. While web, apps, and AI services grew in the spotlight, the firmware-and-board neighborhood stayed quiet. Fewer people wanting to learn means fewer places to teach, and fewer places to teach means even narrower paths into the field. Watching that vicious cycle, I was, honestly, discouraged.

I don't think it's because embedded is a boring technology. It's because it's a technology that doesn't show. Well-made firmware has no presence. The more quietly and reliably a product works, the more the software inside it is forgotten. I understand why a student choosing a career path wouldn't reach for it. I understand — but it still stung.

An ARM Embedded Course, After a Long While

In the middle of that stretch, I took on a two-week ARM embedded course. It had been a long while, so my expectations were high. I had taught a university general-education course called "Software and Artificial Intelligence" since 2022, but that was a broad, shallow tour of how the world works. Spending two full weeks with students on the subject at the very center of my own work is a different kind of thing.

There was also one change I had been carrying around in my head: with the AI and semiconductor boom, the market has been widening. But that was knowledge from news and industry chatter — a story I knew with my head. Whether that change had reached the actual classroom, I had never had a chance to confirm.

A Classroom Has a Temperature and a Mood

To cut to the conclusion: the mood was already good. For two full weeks, the students actively kept up. It wasn't the air of people forced into their seats — it was the air of people who had come to learn this technology.

Stand at a podium for a few years and you learn one thing: a classroom has a temperature. There's the classroom where the only sound is slides being flipped, and the classroom where something is moving back and forth. A teacher can't help becoming sensitive to that difference. This classroom was unmistakably warm. And that temperature didn't cool until the very last day.

But the Strangest Part Was Something Else

The most unusual thing about this course, though, wasn't the level of enthusiasm. It was the composition of the people sitting in the room.

Embedded classrooms have an old formula. Computer science majors on one axis, electronics majors on another, and stretching a bit further, information and communications. For a long time, embedded classrooms were filled from within that range. Those were simply the majors that responded to the word "embedded."

This time was different. Students from all kinds of fields had come to learn this technology. Students from majors I would rarely have met in an embedded classroom before followed the story of the computer's lower layers all the way through two weeks.

Throughout the course, I kept wondering. Why did these people come all the way here?

The Field Has Widened, Top to Bottom

My answer keeps coming back to AI and semiconductors. AI is usually consumed as a software story, but that software ultimately runs on silicon. From data center accelerators to handheld devices to robots, the number of places where compute lives keeps growing — and so does the shortage of people to fill those places. The semiconductor industry's talent shortage is already a story told in numbers, and the roadmap Arm is drawing stretches from chiplets to edge AI to physical AI.

What's interesting is the breadth of that demand. From computer architecture to SoC design to embedded systems — the entire lower half of the stack is being summoned at once. Where "embedded people," "chip people," and "AI people" used to live in different neighborhoods, those boundaries are now blurring. From where I stand, it's a change I still find a little surreal.

The unfamiliar majors in my classroom read naturally in this context. I didn't go around asking each student. But my guess is the path goes something like this: AI enters your field, you get curious about the devices and chips it runs on, and before long you've climbed down to what's underneath — embedded. The story I had only known from the news — "the market is widening" — was something I got to confirm firsthand, in the distribution of majors in a classroom.

Macro trends are things you read with your head. This kind of thing, you read with your body. And signals read with the body stay with you much longer.

Spring for the Tools, Spring for the People

Last spring I wrote a post called The Late Spring of Embedded Tools. I had dissected a chip vendor's new tool and collected the signals that modern software thinking was finally arriving in this slow-moving neighborhood. If what I saw then was spring for the tools, what I saw in the classroom this time was spring for the people.

Did the tools change first and draw the people in, or are the people flooding in and pulling the tools along? I don't know the order. Probably they're pulling on each other. The two things a neighborhood needs in order to change are moving at the same time. That much is clear.

Of course, I'm cautious. A boom is a boom; heat can cool. I know a single two-week course can't speak for an entire market. But for someone who has watched the same neighborhood from the same spot for a long time, small changes like this can be a surprisingly accurate thermometer.

I Hope There Will Be More Chances to Learn

What the course left me with is less an analysis than a wish. I hope students get more opportunities to learn these technologies.

Embedded has an unusually high barrier to entry. Schools rarely teach it, self-study materials are fragmented, and there's the physical barrier of hardware itself. But now the people who want to learn are showing up first. The situation has flipped. Now it's opportunity that's scarce.

Regular courses, intensive ones like this — any form is fine. I want more students from more majors to have the experience of understanding a chip and putting software on top of it. That experience won't turn everyone into an embedded developer. But whatever field they end up in, I believe the gap between people who know how a computer actually moves and people who don't will only keep growing.

The two weeks of lectures are over. Maybe what I needed, having expected so little, wasn't a grand market forecast — it was the faces of students who had come here, of their own accord, to learn.