Ask HN: Why is it so hard to design an open-source smartphone?

IIRC the development budget of the iPhone was around 2 billions and half (did a brief search on google and most sites confirm it was around that estimate), rebuilding a somewhat modern mobile phone (say, circa 2015 specs) without using proprietary components from Qualcomm, Samsung, or even a cheapo chinese foundry like Allwinner would be probably even more expensive.

I didn't know about the size of their budget...

But even for a phone from 2015 there were many models on the market, I don't expect every brands poured that much money in developing a phone. What justifies such a big budget? And what could be a minimal budget for a decent phone?

By decent I mean an average phone: can run most of basic apps, average battery life and pictures quality.

I just did a brief frontend job for a smart-top device, so take this with a grain of salt, but basically when you're planning to release a cheap mobile phone (It was very common until a few years ago to find weird in-house phone brands in European supermarkets) you mainly license a SoC and some sort of development platform from a foundry and that's it.

Most drivers, radio devices, kernel, android releases are those you get at the beginning, you can't really change a thing due to them being closed source or just one step above from being a firmware blob with a few kernel hooks and once your contract expires your devices die with it.

We investigated if we could at least maintain the software on our own (so to not tell some telco clients their streaming devices were gonna die after 4 years) but it was a big fat no from the licensees. And reverse engineering what we already had was forbidden by contract.

In other words, if you want yo build a free phone, you need to build a free radio, a free storage solution, thousands of weird DACs, graphic cards, screen controllers, sensors and whatnot. Making sure you're not infringing on some patents. It's like starting from ground-up. And it's expensive.

A possible solution would be to invest into building into a raspberry-pi binary compatible base platform (and take advantage of their already open sourced drivers), but it was 100% economically unfeasible for a 100-200€ device.