We're not even close to being able to make self-reproducing robots.
Even a simple robot requires a huge supply chain. Take, say, the photoresists used in chip manufacture -- that's an entire chemical industry that you need to box up. That little industry has a bunch of needs on its own. Okay, box those up, too. Repeat. Now let's do insulation for wires . . . or batteries.
It takes a lot more than a village to make a robot, it pretty much takes a country. Probably several, given how suitable raw materials are so inconveniently spread around (and often, hard to find, so I guess we need to add some kind of prospecting facility now).
Short of general-purpose nano-assemblers, which are likely to remain science fiction for quite some time, you'll not put a self-reproducing robot into a probe-sized package any time soon.
The thing is that the first tens of thousands of "self-reproducing robots" do not have to be entirely self-reproducing.
If you can produce 99% of a robot's mass from materials available in space (on the moon or some asteroid), with 1% of the mass consisting of the most complex ingredients provided by Earth (such as the chips you mentioned), you can effectively launch 100x the number of robots from earth at the same cost.
Once you have a couple of thousand relatively flexible robots, you can start to get serious about building specialized labs producing more complex ingredients in space, and then scale up from there.
Even then the task is ridiculously complex, and it seems unlikely that the economics of it would work either[1], but it can nevertheless be much simpler than you make it sound. In any case, we should try to get there because any significant step towards this goal will also make life on earth more robust.
[1] E.g. in the example of chips, good luck getting the masks for advanced designs, let alone all the implicit know-how that goes into a modern fab.
A very important point. Machines which can self-replicate from naturally occurring raw materials pose some daunting engineering challenges.
However, consider how much of our current approach to designing spacefaring robots and their subsystems is influenced by the materials and economic climate we have available here on earth. We deal with severe tradeoffs guided by making robots as lightweight as possible, as reliable as possible (because one failed servo is extremely expensive to replace once you're on mars) and use highly modular components which are well-suited to an expansive global supply chain. What do you design differently when heavy metals are relatively plentiful, weight is mostly irrelevant, scrapping a failed robot is more acceptable and organic substances are scarce? If the parts of the robot are made almost entirely using additive manufacturing processes and circuitry is integrated into the structure you can reduce or remove the need for fasteners and possibly do without conventional insulated sheathing, etc.
The path to self-replication is to reconsider and reinvent a technology stack which minimizes the set of dependencies, and that is a very interesting challenge.
It's hard to project 30 years out. Ask Charles Stross (he's written near-future SF, and -- I'm probably misrepresenting him -- has said he's abandoned story lines because technology caught up with his predictions). Predicting 100 years is pretty much impossible.
But absent "nanotech" or "micro-mechanical biotech" or some similar game-changing advance, self-replicating machines will need to incorporate self-replication of so much of their technology base that they will basically look like factories. Now, there's nothing wrong with a self-replicating factory, but it's not portable and cuddly, and you can't stick it on a rocket and launch it (not in one piece, anyway).
Stross (who is active on HN) explores the subject of abandoning stories in some detail on his blog, e.g. regarding the Halting State series [1]. He also describes here how political uncertainty can kill a story; in this case uncertainty regarding the outcome of the 2014 Scottish Independence vote. Halting State books 1 and 2 are set in a near-future independent Scotland, but were written well before the vote happened. Regarding Halting State 3, he said (2013) "In just two years the map of the Scottish near future will have changed, unpredictably and drastically, from where it is now. I therefore conclude that there is simply no point in my starting to write..."
[EDIT: added quote, made it clear that it was the third book in a sequence that was canned]
It sounds like he set his story too near in the future. Set your story 25-50 years into the future and then you can make reasonable predictions and avoid political things ruining them, for the most part (until a couple of decades have passed). Scottish independence could still happen, just not in the next couple of years. But in 30 years, it's possible. That far out, the state of the entire EU is in serious doubt.
Look at 2001: A Space Odyssey for example. In 1969, it looked like a fairly reasonable prediction of what things would be like in 2001 (42 years away), given the rate of change at the time in aerospace technology. By 1974, it still didn't look too bad. It didn't start looking overly optimistic until probably the late 80s, 20 years later.
Also, if Stross is one of those writers who makes multi-book story arcs spanning over a decade (like Herbert did with the Dune series), that's a sure recipe for total failure when doing near-term sci-fi. Stuff just changes too fast; Herbert's stuff worked sorta-Ok because Dune was set 8000 years in the future (IIRC), but even there one big premise was the idea of genetic memories, which were postulated when he started, but eventually disproven with greater knowledge of genetics, probably before he finished his last book.
This is stuff like Blade Runner worked well: it was a singular story, set about 35 years into the future. At the time, it looked like a somewhat reasonable depiction of 35 years in the future, though rather grim. Of course, now it's almost 2017 and things don't look anything like that, so it's interesting to watch from a historical perspective. It is a little disturbing that they now want to milk it with a sequel after all this time, when obviously things aren't going to look anything like that in 1 year, but I guess I can ignore it like I ignore the Matrix sequels.
We're not even close to being able to make self-reproducing robots.
Even a simple robot requires a huge supply chain. Take, say, the photoresists used in chip manufacture -- that's an entire chemical industry that you need to box up. That little industry has a bunch of needs on its own. Okay, box those up, too. Repeat. Now let's do insulation for wires . . . or batteries.
It takes a lot more than a village to make a robot, it pretty much takes a country. Probably several, given how suitable raw materials are so inconveniently spread around (and often, hard to find, so I guess we need to add some kind of prospecting facility now).
Short of general-purpose nano-assemblers, which are likely to remain science fiction for quite some time, you'll not put a self-reproducing robot into a probe-sized package any time soon.