ARPA-E has a 20% cost-sharing requirement. For example, if the research project has a budget of $1M, the applicant will have to put up $200,000. That may seem like lunch money to policymakers accustomed to trillion dollar deficits, but it is an insurmountable hurdle to startups. Funding opportunity announcements have extremely short deadlines after they are published (3 days), so if you have not been tipped off you won't have time to submit your application. And really you shouldn't bother if you're not rich enough to spare $200,000.
The technology merits don't seem to matter, since DOE maintains no database of technology assessment nor any real outreach. There is no way for the public to engage in a discussion about what gets chosen as a topic for research proposals. ARPA-E does no research, it only decides (arbitrarily) who gets money. Scalability and water impact don't seem to matter either. Case in point: the $1B CO2 underground storage project known as FutureGen 2.0, which can't scale and will push salt into the groundwater. If ARPA-E is what President Obama wants to cite as one of his proudest achievements, he needs to take a closer look, and so should his nominee for DOE.
New technology does not come from the well-connected corporate moochers who scavenge ARPA-E grants. Consider this list of what got funded and who got the money: http://gigaom.com/2012/11/28/arpa-e-backs-66-projects-energy-beets-fabric-wind-blades-dust-devils/
In his farewell letter, ex-DOE Secretary Chu said that some ARPA-E projects had made second base. After four years of a program that was supposed to hit home runs, that should be cause for remorse.