The Quark core 32nm (14nm by end of 2014), x86, core power consumption under 100mW. Assuming performance scales with size/transistor count, that's probably 1/5th-1/10th the performance of Atom (I'm guessing - they claim 400MHz as a max speed at launch).
That's very competitive with ARM at the same size, but ARM doesn't have this "complete computer on a board" at that size (AFAIK).
Knowing the industry terminology helps: https://www.google.co.nz/search?q=arm+system+on+module
Unlike SBCs, COMs [computer-on-module] typically lack real-world ports, and extend their I/O through a connector that supports compatible carrier boards based on standards like COM Express, QSeven, and SMARC. Others use proprietary connectors, often based on a 144-pin SODIMM layout, designed to work only with the company's own carrier boards.
(Although, to be fair I am assuming that the original article is accurate in stating which ports are available on the Intel version)
Further, the OP doesn't really say anything about ports, or availability, and forget about "ports," really. The real difficulty with these is power management. PMICs are getting nice and small, and I'm betting there are a few LDOs on board for generating core voltages, but otherwise all of the real power management will be off board as there isn't much room for inductors on an SD form factor board.
$20 says this is just a cute reference implementation to give to their field application engineers so they can get the OEMs jazzed about Quark.
[Edit: The fact that this is an SD form factor is extra fodder for this argument. This thing doesn't go into an SD card slot, or speak SD, but because SD is something everyone thinks of as "tiny but huge," some marketer somewhere said "make it the size/shape of an SD." For example, the last SOM I integrated (for a computer vision application, built around a 466MHz ARM9 SOC) was of similar size, but of much more useful shape.]
I fear you are correct.
But Intel does seem to be attempting to build RaspberryPi competitors, so maybe this will be in the same line rather than an engineering sample which you can't actually buy.
All that said, that doesn't mean that Intel's technical marketing departments aren't taking risks in areas such as the hobbyist market. Companies as large as Intel are often quite good at this. However, I wouldn't trust the longevity (and therefore future) of anything out of Intel that isn't a high volume silicon product line.
Counterexample: Intel NUC. It's targeted at consumers, and you can actually buy it.
There's nothing stopping you from taking one of these and loading whatever you want onto it, if you can even get your hands on one (you can't).
If it runs Linux, there's a good chance it's open (or could be made open).
I'd like a keychain computer that I could drop next to whatever peripherals I happen to be near.
Of course there are always security problems around using untrusted peripherals. Hell, there are security problems around using trusted peripherals.
That doesn't mean the possibilities shouldn't be explored.
(Incidentally, I don't think this is a very interesting use-case. We can already do that with phones, and I never see anyone doing it. But it isn't security issues that stop people doing it.