Ah, the replicator--or the molecular assembler for the more scientific types. The magic device that everyone thinks will solve all the resource problems. Except it won't. Because simply put, it's magic, not science.
We can look to 3D printers to start with. They are currently quite primitive. They are slow, imprecise, and can only print a handful of materials. I definitely believe that can change--I'm not a pessimist when it comes to technological growth--but slowly. (Does anyone have data on how fast 3D printer precision has increased over time?) It will take many decades to get from sub-millimeter precision to sub-micrometer precision (in my newest timeline, this level is reached in the 2090s, though I am hardly an authoritative figure), and many more to get to sub-nanometer precision. If we get 3D printers that can handle individual atoms are possible by 2062 (or indeed, anytime this century), I will eat my keyboard. If they are common consumer products, I will eat the rest of my computer too.
Consider, too, the mounting engineering challenges that occur with each mounting step, from the macro, to the micro, to the molecular. An especially big engineering challenge rears its ugly head once you get to the atomic level. You see, the process isn't instant. Even Will's timeline describes the process of nano-fabrication as taking "a matter of minutes". And to that, I have this response: atoms react with stuff, and they react with stuff on a timescale far less than minutes. While your molecular assembler is busy manipulating atoms into your desired configuration, the very same atoms will be reacting with each other and everything around them, even the stuff you don't want them to react with. Unchecked, this effect would ruin any object you attempt to fabricate it before you even finish building it.
In my mind, this is quite possibly a mere technical obstacle, not a law of physics. I'm prepared to admit that it can be done eventually, maybe in the 2200s. Maybe later, not before. Once such a machine is built...you still don't have a replicator. It's not like you can just shove random garbage into a magic box, press a button, and get whatever you want. The only remotely efficient way to do this is to have "cartridges" of sorts for each element you need, and those cartridges would cost money. You could have a machine that takes a mass of random junk, finds the right atoms, and sticks them into place, but that would obviously be far less efficient, and likely take centuries more than a "vanilla" molecular assembler. The files containing atomic assembly data also wouldn't be free, by the way. This is called an information-based economy.
I am, of course, aware of atomic transmutation. But using that for any practical form of is centuries more advanced than a molecular assembler. Look how hard nucleosynthesis is currently: it took millions of dollars and years of effort to create seven friggin atoms of tenessine. Yes, that is a super-heavy element, but we would also have to create a trillion trillion times as many atoms, and for a million times lower cost, for this to be economically practical. I can accept many things, but I can't accept that synthesizing light elements out of hydrogen will be thirty orders of magnitude more efficient than synthesizing super-heavy elements. Beyond that, nucleosynthesis-based manufacturing would require a masterful control of nuclear reactions. You would get the same problems as atomic manipulation, but at the sub-atomic level, where everything is orders of magnitude more difficult. Instead of controlling chemical reactions, you would have to control nuclear reactions.
So there you have it. Molecular assemblers. Doable? Yes. Good for extreme precision manufacturing? Yes. Easy? No. A magic box to make whatever you want for free? No.