As mentioned, the same is said of the main product line these days. There they've gone a long way up in TCO without an equivalent feature/performance differentiation over the competition to show for it. They can't all be the non-competitive ones the average non-k12 student is supposed to avoid looking at each time the conversation is brought up. Either the main line needs to become hyper competitive again or the accessory lines can take the role, just not "neither".
I think the Pico 2 and maybe compute module variant (depending on sub-niche use case) are still competitively interesting in their own right, but feel the rest are largely help up by the brand name at this point and I wonder how much of the revenue those add up to (Pico in particular, since its market is in the cup of coffee range).
Of course the brand value will hold things for many years to come (how many have 3 pis in a drawer just because the new one comes out and folks still buy it on name before thinking if it's worth it?) so they've got plenty of time to strike gold again for now.
The Pico/RP2040/RP2350 ecosystem is more competitive because it's a distance-to-the-metal thing.
At this point, most "mainstream" Pis are being used with off-the-shelf software or at least an off-the-shelf OS with custom userland code, and the technical details of the Pi are black-boxed away. So any competitor just needs to get you to the same basic "here's a Linux distro with the common packages" to get to a basic product parity, and then can differentiate with more/better/cheaper. This is especially easy when you can target a vertical specifically, like the "router boards" that come with an OpenWRT package.
The Picos are being programmed at a low level. If you want to swap in a STM32 or CH32V, you're a lot more concerned about "are the reference docs available and accurate", "will there be a reference to my specific weirdnesses on Stack Overflow", and "do the dev tools actually work." From that perspective, the RP products are industry leading, at least to a "Nobody ever got fired for buying ~IBM~ RP2040" level.
STM32 would be the IBM of the embedded world, but that doesn't invalidate your point; RP MCU's are documented well enough and stable enough for professional work.
> There they've gone a long way up in TCO without an equivalent feature/performance differentiation over the competition to show for it.
Yet they’re still selling out quickly and it can still be hard to find the model you want without shopping different distributors.
I think people like you who comparison shop based on the specs in a table don’t realize you’re simply not the target market. The target market will take the better ecosystem, support, and documentation even if it comes with less performance.
The people who want the fastest SBC and don’t mind spending a day or week chasing the right kernel fork to solve their problem are not the target audience.
I think people tried out the Raspberry Pi 4 because it was just powerful enough to run desktop and there weren't cheap desktop options. Now there are plenty of cheap desktop options and while Pi 5 is more usable, it can't compare.
But the Pi was popular before that for fun things. It still compares well for fun things unless trying to do AI. For me, the Pi 5 isn't that interesting because I don't need the performance, and Pi 4 or Pi 3 will get the job done without the power hassle.
it's a lot easier to buy additional gadgets when you can assume or when they explicitly declare that they're usable with the pi 5, and the time saved from "oh just install these commands" documentation is easily much greater than the cost for the same price. And the performance isn't the point either, the tinkerability and time spent is.
And I think they'll continue to sell very well from brand name alone. That won't last forever, but it's not going to disappear next year either. They've had pretty flat unit volume the last 3 years, it's not about to nosedive out of the blue the same as volume is not about to continue growing the same as it used to (even though they launch more and more product lines).
The differentiator for the Pi line was support for the capability at the price point, but the competition for the mainline Pis is no longer "random ARM boards with no driver updates", so no longer is that support actually the differentiator for the Pi either. With the lowest cost Pi 5 model, $50 gets you a bare 2GB RAM board with no power or storage and you're still left with the bespoke ARM OS images and binaries to deal with. For the price of higher end models you can just get a complete standard x86 PC which happens to run better. That latter bit about the spec sheet is a bonus, not the main change. I.e. the target market shifted from "hot damn, I can run Linux at such a low price and wattage point while not worrying about support???" to "I want to spend my weekend tinkering with a Pi". The Pico 2 remains decent for the general market though. At $5 it's a well supported MCU with decent kit and a USB interface if you need to have that on a computer.
It'll also be interesting to see how much education even remains the mission now that they've IPOd. E.g., the mission statement on the investor relations page is:
> Raspberry Pi’s mission is to put high-performance, low-cost, general-purpose computing platforms in the hands of enthusiasts and engineers all over the world.
At the end of the day though, we could talk for days about how it must be one way or the other, but the only way to see what will actually happen is to wait 5-10 years. That reminds me, there is a regular HN "predictions for the next decade" kind of thread every turn of the decade and we're closer the next one rather than the last one already!
I think the Pico 2 and maybe compute module variant (depending on sub-niche use case) are still competitively interesting in their own right, but feel the rest are largely help up by the brand name at this point and I wonder how much of the revenue those add up to (Pico in particular, since its market is in the cup of coffee range).
Of course the brand value will hold things for many years to come (how many have 3 pis in a drawer just because the new one comes out and folks still buy it on name before thinking if it's worth it?) so they've got plenty of time to strike gold again for now.