First, suppose we have a drug candidate. Great. This would take many, many years of study of neurons/brain tissue cultures using in vitro techniques (petri dish, etc. out of the native environment). These are great and relatively cheap, but it's difficult to isolate good drug candidates. Eventually, after high throughput screening or whatever other research method is used, a candidate is found. This would (realistically) take many years and require many millions in funding.
Then, once you have this in vitro drug candidate, you have to start testing it in animal systems, starting with simple mice. Animal studies are expensive, because the organisms are essentially disposable. These sorts of studies for neural problems are even more expensive because the animals need to be observed carefully using various research techniques for very extended periods of time, on the order of months to years (since this is the timescale on which most neural disease work). This is obviously both expensive timewise and monetarily.
So our theoretical candidate works in simple animal and in vitro systems. Great! There a lot of drugs that get this far and then utterly fail on human subjects. Assuming that it passes FDA standards for human testing (which, at this point, seems fairly reasonable), human testing will start. The main purpose of Phase I trials is to determine toxicity of the drug in human systems, and to see if it has any significant effect. Let's say ~40 patients are being tested. These 40 (lucky? unlucky?) patients will take the drug, and then be investigated using a variety of techniques to judge the efficiency. One of the standard techniques in neural imaging is the MRI scan. The typical cost to the company carrying out the test is around $3000+ USD per scan carried out. Each patient will need to be scanned many times (let's say 10) over a period of time to determine whether or not the drug had any affect. Compared to other techniques where simpler blood tests or post-mortem tissue analyses can tell the story, this is very expensive. And this is just phase I. Phases II and III expand the number of patients into the several thousands. The cost of drug development for one potential compound for so long (the timeframe can be order of 15+ years) for a very, very small chance (due to the complexity of the field and the few successes so far) of a huge earning is simply unjustified for many companies. Earning back hundreds of millions of dollars of losses back must be done by passing the cost along to the customer.
Note that the patent lifetime on drugs is very short. Nominally, the compound is protected for 20 years. But this is applied for before any clinical trials start, so the effective lifetime is significantly shorter. For a neurodegenerative disease, the effective lifetime of the patent could very well be much sorter than the more "usual" effective lifetime of 7-12 years [1].
Also, as an extra. Organic synthesis is hard. I mean, really, really hard. You have to design a good synthetic pathway, then make sure it works, then scale it up from working in a fume hood, to working in large chemical reactors that can produce it on an industrial scale. I'd like to point people at the synthesis of Tamiflu, the antiviral used against various influenza strains [2]. These are long pathways and accordingly tend to have lower total yields. Drug synthesis is really hard, regardless of the cost of R&D of the drug.
TL;DR: Finding drug candidates is really difficult. FDA clinical trials is really difficult. Making the drugs themselves is really difficult.
[1]: http://en.wikipedia.org/wiki/Generic_drug [2]: http://en.wikipedia.org/wiki/Oseltamivir_total_synthesis