In the labs of young startups and universities, researchers are fine-tuning groundbreaking fuel technologies and "out-innovating" other countries in the race to make clean energy.

That, at least, was the picture the Department of Energy wanted to convey at this week's third annual ARPA-E Energy Innovation Summit. The event is an expo of high-risk energy inventions backed by DOE, held mainly for would-be investors.

This year it doubled as an appeal to Congress to save ARPA-E, DOE's Advanced Research Projects Agency-Energy, which is running short on money.  

At the Gaylord Convention Center near Washington, D.C., where the summit was held, Silicon Valley startup Envia Systems unveiled an electric car battery that costs half as much as lithium-ion cells and travels three times farther on a charge. California's Makani Power test-flew a mockup of its airborne offshore wind turbine, which generates electricity at low wind speeds while tethered to land via cable. Michigan State University showed off its wave disc engine, an alternative to the internal-combustion engine that increases mileage to 100 mpg in gas-engine and hybrid cars, and produces 90 percent fewer emissions compared with today's models.

All three projects received money from ARPA-E, once pitched as a "Marshall Plan" for energy. Since 2009 the fledgling agency has awarded more than $520 million in research grants to about 180 projects that promote American innovators in areas ranging from more efficient solar cells and air conditioners to grid storage for renewables.

The goal is to form a incubator for transformative technologies, so startups and labs can attract scale-up venture capital and private equity funding. DOE says 11 ARPA-E projects have lured a combined $200 million in private investment, five times what DOE put into them.

Yet the agency, which was created by the 2009 stimulus, is in a pickle. ARPA-E got a $400 million jumpstart in 2009—and it received a combined $455 million in the 2011 and 2012 budgets—but its funds are quickly drying up.

While Pres. Obama requested $350 million for ARPA-E in his 2013 budget proposal, it comes at a time when Congressional Republicans have promised to slash the amount of taxpayer dollars going into new energy technologies. Many soured on such programs after Solyndra, a solar startup backed by a $528 million DOE loan, went bankrupt. This week news broke that Abound Solar, a Colorado solar firm that won $400 million in DOE loan guarantees, was firing 280 people and abandoning plans for a second factory. Some analysts see signs of trouble for two other loan winners, Fisker Automotive and SoloPower.

At the summit, DOE brought in high-profile personalities to shore up support for ARPA-E, including Bill Clinton, Bill Gates, former Walmart CEO Lee Scott and Susan Hockfield, president of the Massachusetts Institute of Technology. All argued that projects nurtured by ARPA-E—which is modeled after a Department of Defense program that helped create the Internet and GPS systems—could change the global energy landscape if given a significant amount of time and funding.

"It would be a terrible mistake" not to "invest even more money in ARPA-E," said Bill Clinton during a keynote address Wednesday.

It's a tricky sell. Even ARPA-E's staunchest proponents admit that many of its projects could flop. Last month DOE said that nearly 35 projects, about 20 percent of the program, are struggling, the New York Times reported. Since September DOE has canceled six ARPA-E projects that won a combined $14.1 million. (The Treasury got $3.7 million of that back, a DOE spokesperson told Bloomberg.)

None of this surprises Bill Gates and his fellow ARPA-E supporters. "The failure rates are going to be well over 90 percent," Gates told the summit on Tuesday. "This is a very complex set of technologies, so we need literally thousands of companies trying these technologies so we can get the 10 or 20 that [make it.]"

In reality, even if Congress funds ARPA-E, it's private investment that matters most in the end. Investors can bankroll the factories that "moonshot" technologies need to deploy, and their appetite for these projects appears to be waning, the Wall Street Journal reported on Wednesday.