The previous +1 answer gives the correct economic mechanism but let me give you some case specific answer here.
First of all there actually already is another drug that targets the same disease called Spinraza (see here) with cost of \$750,000 for the first year and then \$350,000 per year after that, so about \$4 million a decade. Assuming people with this deasease can live up to 70 years (which would cost patient \$28 million over life time) that means the Zolgensma drug is already $92.5\%$ cheaper than their competitors over patients lifetime (although the figure is less impressive if this person dies earlier - but even at 10 years it is already almost half the price of the other treatment). Sure this does not answer question why it could not be even more cheaper - but it already puts the high price tag in some perspective.
Given that these more expensive therapy exists I assume that the market for it wont be 0. Even if the cost is high it can be manageable with insurance that spreads the high cost over large number of people. I doubt most patients will pay the price out of the pocket.
Second, the article you cited says that this drug was developed for a rare disease that affects only about 1 in 10000 people, so it is not like this drug can be sold by thousands.
Let us just do some back of an envelope calculation. These gene therapy treatment is particular valuable and applicable to those diagnosed in early childhood. According to the Our World In data the population increase in 2019 was about 80 million. Now many of these people will live in countries without sufficient infrastructure/supply chains etc to administer any treatment. Especially in sub-Saharan Africa there are issues even with giving people cheap deworming pills due to conflicts/anarchy, people living in rural areas without clinics, lack of education etc. so no matter what is the price it won't be accessible to about $33\%$ of world population. That gives us about 50 million. Since the disease will affect about 1 out of 10000 people that means yearly the market for the drug is about 5000 people.
Next, we actually know exactly in this case how much the medicine costs because it was not developed by the company but acquired from another company reportedly for $8.7 billion from US firm AveXis. Furthermore, patents for drugs last 20 years from the day they are patented and since the FDA approval process takes some time and there is always a threat of a new competitor coming up in with a new drug it is not reasonable to say company will have all 20 years to recoup costs. Let's for a sake of argument say that it will have 10 years. Furthermore, these are just development costs. As far as I understand gene therapy is not as easy to manufacture as regular pill (but I am not medical professional), plus there are all other associated with running the business but let us forget about those for a second.
Well that gives us the total market for this drug over 10 years to be about 50000. Now we need to spread all the 8.7 billion over 50000 people. Hence to break even (just on development costs and nothing else) the drug would have to be priced at \$174000. Again this is just development cost so I would guess that at your proposed \$200000 the company would still not break even.
In addition any company needs also some healthy profit because R&D is risky business any wins have to pay for all failures to successfully develop drug. Even in cases like here where R&D was done by different firm there is still an issue that you never know when the new competitor arrives. If we would change the above assumption that the company will enjoy monopoly for 10 years to just 1 year because maybe better treatment will come next year suddenly just to break even on development the treatment would have to cost \$1740000 which is much closer to current price.
In fact according to Institute for Clinical and Economic Review, that took into account wide range of factors the new gene therapy would be justifiable at cost \$1.1 million to \$1.9 million per treatment. Now this is still lower than the \$2.1 million per treatment but less so.
Now this answer is actually complementary to the answer of Bayesian above since undoubtedly thanks to their monopoly on the drug the company charges more than it would if there would be more competition and this mechanism is correctly described by Bayesian, but the point is that even without this issue assuming that price would have to cover the development cost etc it would still have to be priced around million.