The state of molecular manufacturing today can usefully be compared to the state of nuclear bombs in 1940.
In 1940, it was known to a few phyisicists that some heavy isotopes could be split, and could in theory create a chain reaction. Plutonium had not yet been discovered. A chain reaction had not yet been demonstrated. Einstein had already written his letter to Roosevelt, in which he used lots of phrases like "it may become possible" and acknowledged that an atomic bomb might well be too heavy to be delivered by airplane.
At the end of 1942, the first chain reaction was created. It required over 70,000 pounds of uranium, and 771,000 pounds of ultrapure graphite, to make it happen.
On Dec. 28, 1942, Roosevelt created the Manhattan Project.
One of the problems to be solved was separating the isotopes. They tried three methods, and made all three work for at least some stages of enrichment, though they eventually settled on gas diffusion. That required a six-story building covering 43 acres -- half a mile long -- full of 4,000 diffusion stages, without a single leak in the pipes.
In mid 1944, they discovered that their original "gun" design wouldn't work for plutonium, so they designed an "implosion" system. In late 1944, quantities of fissionables material started to arrive at Los Alamos.
On July 16, 1945, the first atomic bomb was exploded.
Molecular manufacturing requires the use of sub-micron engineered machinery to build more sub-micron machinery. Biology does something similar, except that it's based on evolved complexity rather than engineering. Large-scale proof-of-concept systems have come close to building duplicates of themselves from simpler parts. There aren't any theoretical fundamental reasons why engineered productive nanosystems can't work. But there aren't any detailed designs yet, either.
There appear to be several ways to build the first nano-building-nano system: several different materials might work, and there are several basically different ways to handle them. No one knows yet which way will turn out to be easiest. And no one knows how easy that will be. Building the first tiny nanofactory might require a thousand people using a thousand scanning probe microscopes for several months -- an effort substantially larger than most companies would be willing to fund, but substantially smaller than what was required for uranium separation. Or it might only take ten million base pairs of DNA synthesis -- which would cost just a few million dollars.
In addition to cost, another major question is the calendar time required for research. Of course, to completely understand a chemical system would require many years. But molecular manufacturing doesn't require complete understanding -- just enough to do a limited number of synthesis reactions reliably.
If a nation or large company wanted to develop molecular manufacturing very quickly, and was willing to spend a lot of money on it, they would not use an academic research model. They certainly would not use a formal peer review system! Instead, they would launch several approaches in parallel, perhaps with several teams of experts placed explicitly in competition with each other. Within six months, a small team of experts could give a preliminary ruling on a material/method combination. (Method refers to the way reactions are done, such as machine-phase chemistry, position-controlled solution chemistry, or guided assembly followed by crosslinking.) A diverse, energetic, well-managed, well-funded effort could probably investigate twenty or fifty such combinations within the first year. After that, any approach that looked halfway good would be funded in parallel.
Once a material and method was selected, how much time would be required? The material's chemistry would have to be researched via massive simulation (which could use computers prepared ahead of time) and experiment (using fairly standard lab techniques). Its mechanical properties, including molecular mechanics, would have to be investigated. Basic machines would have to be designed. All these could take place in parallel, with the answers being increasingly refined. I'd expect that crude numbers and placeholder designs could be available within another year, and sent to the bootstrapping team and the mechanical design team. And so on...
Of course, this depends on three things. First, the willingness to spend large amounts of money to get answers fast. This probably means a military program.
Second, an attitude of "Don't tell me why it can't work, tell me how to do it!" Or as inventor Jacob Rabinow said (quoted in Creativity p. 69), "There's one other thing that you do when you invent. And that is what I call the Existence Proof. This means that you have to assume that it can be done. If you don't assume that, you won't even try. And I always assume that not only it can be done, but I can do it."
Third, it depends on there actually being a solution to find. But I don't know of anyone who's taken a decent look at molecular manufacturing who thinks it's actually impossible. The questions now are how hard it will be to develop, and how high performance the manufacturing systems and products will be.
Do I expect it to be developed in five years? No. If the U.S. had a "Nanhattan Project" going today, I'd give it maybe a 30-70% chance. With no evidence of such a project, I'd put it closer to 5-10%. But there's been such a pitifully small amount of investigation so far that it's quite possible there's a relatively easy way to do it that no one has noticed yet. And it's even possible that the entrenched skepticism in the U.S. will lose traction, making it possible for people to publicize research on it; if an academic discipline were ever allowed to form, it probably wouldn't take long for interest in the technology's power to snowball.