Almost every commercial hardware product shipping today sits on a foundation somebody else built in public. Linux runs the Android phone in your pocket. Linux runs the New York Stock Exchange. ARM processors that started life in mobile reference designs now run satellites, scientific instruments, robotics platforms, and most of the embedded products an OEM is likely to ship this year. The hard stuff under the hard stuff usually got solved decades ago, by people who gave it away.

Hardware leaders are already using open source hardware. The harder question is what to do with the layer of abstraction sitting between their product and the foundation underneath it.

Why a Research Lab Built Its Own Audio Computer

Andrew McPherson is the cofounder of Bela and a professor at Imperial College London, where he runs the Augmented Instruments Lab. Bela started inside that lab around 2014 as a tool for a single research project. Andrew was building an experimental instrument called the D-Box and needed real-time audio processing with end-to-end latency around one millisecond. A laptop cannot do that, neither can phones. The hardware to run audio code at that kind of latency had to live closer to the metal.

So Andrew built it. The technical foundation was a BeagleBone Black, which at the time was the only board capable of meeting the latency requirement at a price a research lab could afford. The instrument got built. The platform underneath it turned out to be useful for a lot more than the D-Box. Other researchers in the lab started using it. So did collaborators outside the university. Then teaching faculty started using it for digital signal processing courses. By 2016 the team had run a Kickstarter and turned the platform into a company.

Bela’s newest hardware, the Bela Gem, runs on the PocketBeagle 2, released in the past year. Quad-core, several times faster per core than the original.

“We had a debate for a lot of years around how much we should prioritize looking into the directions that increase CPU capacity. For a lot of years I really thought it doesn’t really matter. People weren’t really hitting the limits of that. So you don’t just do it just for the sake of specifications. You do it because people have real practical needs.”

What changed was that audio developers started building patches and signal chains that hit the limit. PocketBeagle 2 arrived around the same time the demand did, and Bela Gem followed.

What Beagle Was Built To Do

Jason Kridner is the founder of BeagleBoard.org, the nonprofit foundation behind the BeagleBoard family. He started Beagle while he was at Texas Instruments in the late 2000s, when ARM processors lived almost entirely inside cell phones and the open source software world had no capable, affordable hardware to run on. The bet was that giving the open source ecosystem a real board to use would produce a category of innovation that proprietary reference designs would not.

That bet paid off in directions Jason did not predict. Jet Propulsion Lab uses Beagle to prototype Europa rovers. Fermilab standardized on Beagle for data acquisition. Someone built the world’s first cheeseburger robot on one. Most of those projects came from small teams or individual builders who could afford the cost of being wrong because the cost of starting was close to zero.

Open hardware lowers the price of finding out whether something will work. That is the price that determines whether a small team gets to try at all.

The Veneer Problem

Jason has a sharp opinion about how to build on top of someone else’s open work. The layer underneath your product should remain visible to anyone who wants to look at it.

“When you start hiding it under your own veneer, that veneer gets very fragile. We’re building on top of Linux, we’re building on top of other open source technologies. And we try to expose those as directly to users as possible. When you try to abstract the world away, it’s not necessarily helping people build up an understanding of the world. There’s a right way to make it easy to use, but not hide the details that you need to actually go further.”

Bela embodies that principle in product form. A Bela user who wants to write audio code in the browser-based IDE and never think about what runs underneath can do that. A user who wants to peel back the abstraction, look at the TI processor data sheet, write directly against the Linux real-time extensions, and access the Beagle pin headers can also do that. The simple path and the deep path coexist on the same board.

The veneer point extends beyond Bela. Any product built on top of someone else’s foundation makes a choice about how thick to make the abstraction. Thick veneers create initial ease and long-term fragility. Thin veneers make the foundation visible, which means experienced users can extend the product in directions the original team never anticipated. That extension is where most of the interesting work happens.

What Open Hardware Looks Like From a Closed-Source Manufacturer

Amtech is a close-sourced shop. But on that pays close attention to the open hardware ecosystem. At Amtech, we build proprietary products for OEM customers. Many of those customers, especially smaller venture-backed hardware teams, are trying to compress the time and capital between an idea and a working prototype. The capital required to design custom silicon, lay out a board around it, and validate the design can be large. Open hardware ecosystems shrink the front end of that work to something a small team can absorb. The ideas that make it through to production benefit from a foundation that the team did not have to build themselves.

Our CEO Jay Patel makes the broader point this way: open hardware is a phenomenal enabler for economies and industries and creativity. Without it, innovation would not happen at the scale or pace it does. That is a closed-source manufacturer arguing for open foundations on operational grounds. Customers who can prototype faster generate more programs. More programs generate more manufacturing demand. The ecosystem feeds itself.

What the Brand Question Misses

The objection from inside a hardware company is usually some version of: but if we publish our designs, anyone can copy us. Jason think that’s pretty short sighted.

“The idea that you make a board and somebody can’t figure out how to make a similar board is largely unreasonable in this world. It’s not so hard to reverse engineer if there’s an opportunity for it. Giving away the details of your board is not dramatically changing the barrier to entry for professionals.”

The barrier you think you are protecting with closed designs is mostly imaginary at the level of an experienced competitor. Closed designs make it harder for less experienced builders to learn from you, less likely for a community to form around your platform, and less defensible when someone with more significant resources decides to enter your category. The defensible asset is brand and accountability. People know what they are getting from Bela. They know who is responsible when something goes wrong. They know where to find help. That is what you protect, and a published schematic does not weaken it.

Where This Leaves an OEM

A hardware leader looking at an open source ecosystem is looking at a shortcut around the work that’s already been done. The Linux kernel, the toolchain, the reference designs, the community of people who know how to use them. The first decision is whether to skip that shortcut for some philosophical reason or take it. Almost everyone takes it.

The second decision is harder. How thick do you make the layer between your product and the foundation? A thick layer feels safe but often produces fragile products. A thin layer feels exposed but can produce products that experienced users can extend. The teams building the most interesting things in hardware right now are instrument designers, lab researchers, satellite builders, and robotics engineers, and they keep choosing the thin layer. They want to see the foundation underneath. They want to know what they are building on, and they want to be able to go deeper when the project asks for it.

The closed-source shop is one of the things the open ecosystem makes possible. Most hardware companies shipping something interesting in the next five years are sitting on a foundation they did not build. The ones who let that foundation stay visible to their own users will outlast the ones who paint over it.