I think ARPA-E is a great idea. Civilizations almost always under-invest in basic research. Applied research for which there is a near term possibility of economic payoff tends to be reasonably well supported by the market if institutions such as venture capital or corporate R&D are well-developed. However, increase the expected timescale for economic return by a few additional years and suddenly the amount of non-governmental funding available drops precipitously.
The early days at ARPA-E were pretty insane. Its first couple of employees had to put out its first solicitation, and it was inundated with 3700 applications for its first 37 grants, which crashed the federal computer system. But they attracted an absurdly high-powered team of brainiacs: a thermodynamics expert from Intel, an MIT electrical engineering professor, a clean-tech venture capitalist who also taught at MIT. The director, Arun Majumdar, had run Berkeley's nanotechnology institute. His deputy, Eric Toone, was a Duke biochemistry professor and entrepreneur. Arun liked to say that it was a band of brothers; I like to think of it as a $400 million Manhattan Project tucked inside the $800 billion stimulus.
... So far, more than a dozen ARPA-E-funded companies have already attracted follow-up venture funding. They're very excited about 1366 Technologies, which has developed a new solar manufacturing process. Basically, instead of slicing silicon ingots like salami, which is a difficult way to make wafers and wastes a lot of silicon dust, they're creating the wafers directly from liquid silicon like pancakes, which could cut the price of solar panels by a third. The other big winner so far is Envia Systems, which has developed the world's most powerful lithium-ion battery; it could slice $5000 off the cost of the second-generation Chevy Volt. But there are all kinds of exciting projects: lithium-air batteries that could put lithium-ion out to pasture someday, wind turbines shaped like jet engines, electric transformers the size of a suitcase instead of a kitchen, laser drilling technology that could cut costs of geothermal wells as well as petroleum wells. We'll see what pans out.