In just a few years, most Americans will abandon car ownership in favor of on-demand self-driving cars, think tank RethinkX predicts in a major new report. It’s not just that autonomous Ubers and the like will be more convenient than regular cars (summon them with an app, don’t worry about parking, don’t worry about driving) but, also, that they will be much cheaper.

Relying on on-demand services will cost a projected $3,400 annually by 2021. Buying a new gas car, conversely, will cost $9,000 annualized. (Buying an electric car narrows the gap a little; using an old car narrows the gap even more.)

Below is a chart showing how much cheaper Transportation-as-a-Service (TaaS) will be than buying a new gas (ICE) or electric (EV) vehicle — or even using an existing gas vehicle.

For most Americans, that will be an easy choice. That’s why 95 percent of U.S. passenger miles traveled by 2030 will be in self-driving rentals, RethinkX predicts.

Who won’t switch to self-driving rentals? Rural consumers, whose communities “may not have the population density to have high enough demand to attract a critical mass of TaaS vehicles.” Tech laggards “who will not switch to TaaS for a range of personal reasons, including dislike of change, distrust of new technology, and perceived loss of personal freedom.” And the very rich, “who are not motivated by road travel economics, despite the scale of the savings that TaaS offers.”

What happens if so many people switch to on-demand cars? Some pretty cool stuff. The report predicts beneficial consequences, including $1 trillion in savings for U.S. households over ten years; $1 trillion in GDP growth from increased productivity; 200 million fewer passenger cars on the road by 2030, freeing up tons of land for more productive use; and more.

Photos via RethinkX, Flickr / Nicholas_T