Cars have not been good for the environment, to put it lightly. Transportation accounts for 28 percent of US greenhouse gas emissions, and light-duty vehicles for more than half of those. Someday, self-driving cars will appear widely in the US. Wouldn’t it be nice if they also helped reduce greenhouse gas emissions?

Trouble is, making an electric car self-driving requires tradeoffs. Electric vehicles have limited range, and the first self-driving cars are expected to be deployed as roving bands of robotaxis, traveling hundreds of miles each day. Plus, the sensors and computers onboard self-driving cars suck up lots of energy—not great for range, either.

New research suggests that the tradeoffs for electric autonomous vehicles aren’t as painful as once thought—and indicates that AVs, whenever and wherever they show up, could contribute to the green-ing of the global car market.

In a paper published in the journal Nature Energy last month, researchers from Carnegie Mellon University project the potential behavior of self-driving cars in cities and suburbs. They find that certain aspects of autonomy do drain car batteries, but smart software and hardware tweaks should make fleets of battery-powered self-driving cars very possible.

“A bunch of commentators used to suggest the first AVs might have to be gas hybrids,” says Shashank Sripad, a PhD candidate in mechanical engineering at Carnegie Mellon who worked on the paper. “But we believe that, if we want to do electric vehicles, autonomy will be compatible with it.”

Automakers differ on whether to power their first self-driving vehicles with electricity. The intra-industry divide is a reminder that autonomy is both an ambitious research project and a potential multi-trillion-dollar business, and that different players see different paths to market. The ideal self-driving business model, in other words, is far from settled.

Take Ford. The company aspires to transition to battery-electric self-driving cars eventually, says Dan Pierce, a spokesperson for autonomous vehicles at the Detroit automaker. But if Ford hits its deadline of launching an autonomous vehicle service in 2022, it will do so with gas-electric hybrid vehicles.

The WIRED Guide to Self-Driving Cars

How a chaotic skunkworks race in the desert launched what’s poised to be a runaway global industry.

For now, Ford’s testing shows that more than 50 percent of a battery-electric vehicle’s range would be sucked up by the computing power demanded by self-driving software, plus the air conditioning and entertainment systems needed to keep passengers comfortable. Ford also believes that the fast charging required to run a fleet of self-driving cars would degrade the electric battery too quickly. Finally, the carmaker has crunched the numbers, and thinks it would waste too much time charging electric vehicles—time otherwise spent transporting paying passengers. “We need to find the right balance that will help develop a profitable, viable business model,” Pierce says.

Hometown rival General Motors, and its self-driving-tech subsidiary Cruise, have other ideas. Earlier this year, Cruise unveiled a six-seat, toaster-like electric vehicle it says will be the foundation of an eventual self-driving ride-hail service. Until then, Cruise plans to rely on an all-electric Chevy Bolt, which it has tested in San Francisco. (Cruise once intended to launch that service in the city in 2019; the deployment has now been delayed indefinitely.)

Rob Grant, the company’s vice president of government affairs, says Cruise wants to make charging more convenient by building more charging infrastructure in city centers; Cruise says it owns 40 percent of the fast chargers in San Francisco. He also sees the focus on electric as a valuable part of Cruise’s pitch to governments, particularly in California. “A lot of regulators are focused on sustainability,” he says. “Our mission dovetails with the state’s mission of reducing greenhouse gas emissions.”