I think it's important to note that comparisons to "the physical hardware costs $x" or "a start-up could have done this project for $y" are completely invalid. This is an easy trap for anyone who has ever tinkered in their basement or hacked away at code can fall into. In Real Life (TM) no massive integrated system is built by MacGyver on a $5 budget.From the engineering perspective, Federal Online exchange marketplace is a massive and complex undertaking. It is a massive project with strict deadlines. No one with a software engineering background and a bit of common sense are surprised with the problems plaguing the system.
First, recall that this is a government project. There are inefficiencies and bureaucracy. Preferred suppliers. Yes, even political sweetheart deals. Percentages to the raw cost get added at every level; imagine 5 to 10 levels of subcontractors. Second, this project deals with healthcare data. The sheer amount of laws that govern the usage of this data are mind boggling. Not to mention every state has a different set of rules. This complexity is incredibly high, inefficient, and requires many, many domain experts as compared to a standard "big data computer, turn the crank" system. Lawyers, Health Policy people, Doctors, Bureaucrats in health, etc.
Third, this project has insurance companies. Insurance companies have an even more complex system of actuarial science and secret sauce that they use to package policies. Think about how long it takes an agent to get a single quote together for you. A half hour on a good day, and a week on a bad one. Now imagine every insurance company in your state having to produce a quote, using a different system of their computers and algorithms, and black magic based on this juggernaut of a government system. At this point, we can finally add in the cost and complexity of the physical system and work to physically build/program it. It's probably less than 1/8th the total cost.
Am I saying it is optimal? No. Am I saying it is the worst usage of government spending ever? No.
Finally, some perspective. Government spending is somewhere around 3.8 trillion for 2013. $634 Million is approximately 0.01% (yes, that is percent) of all government spending. Saying that a theoretical "startup" could do it for 1/6th is like taking a bucket of sand from the Sahara desert and claiming success in a terra-forming project.
TL;DR Government Bureaucracy coupled with healthcare data coupled with insurance companies coupled with a physical project is very expensive. It's not optimal, but when has government funding ever been "highly efficient" in the way it produces goods? Reddit
The system will get stable at some point. Right now ignore all the political talk and be grateful that a somewhat functioning health care program is law in the US.
And this comment is also worth reading in full.