This year we’ve seen an uptick in U.S. car travel, likely triggered by cheap gas prices. The shift is notable given the general trend for a decade — per capita Americans are driving less.

While driving is up in 2015, it’s nothing compared with the explosion in vehicle traffic we’re in for by 2050, according to a new report from KPMG. It estimates that between population gains and the popularity of fully self-driving mobility services, we’ll see the total number of vehicle miles grow by 1 trillion. (Half of the 1 trillion it attributes to population growth.) For perspective, U.S. residents drove 3.1 trillion miles in 2014.

KPMG expects this growth to come from trips taken by the very young and very old, who can be immobile only due to their inability to drive. By having access to a self-driving shuttle, a world of opportunity would open up.

“In many ways we may be way underestimating this phenomenon,” said Gary Silberg, one of the report’s authors. “We didn’t even touch the middle [demographics]. We didn’t talk about people who might use mobility on demand because they’re just going to change their habits.”

The report notes that if we end up with empty self-driving vehicles on roads — which would seem likely as cars drive to their next customer or retreat to an affordable parking zone far from an urban core — the increase in mileage could be as large as 3 to 4 trillion additional miles. This would effectively double the traffic on American roads.

“Do we have the proper infrastructure for this? I think that’s a fair question,” Silberg said. “Once you get over 1 trillion or 2 trillion, 3 trillion — the urban planners and all the transportation people better start thinking about this now.”

KPMG did focus groups of about a dozen people in Denver, Chicago and Atlanta to gauge interest in autonomous mobility services. Those results are not statistically significant, but they jibe with common sense.

Over the summer I spoke with Sevil Yasar, a Johns Hopkins researcher focusing on people with dementia and memory problems. When safety calls for it, she has to tell patients that they should stop driving, and it’s a conversation that almost never goes well.

“If there would be a self-driving car where you tell the GPS where you want to go, and it can take you there, I think that would be a fantastic possibility,” Yasar said.

For an elderly person, losing driving privileges begins a “vicious cycle,” according to Yasar. They’re likely to become withdrawn, less social and remain at home. If every senior citizen who had to give up their license could tap a button on their phone and have a self-driving vehicle appear at their front door, you could count on them being more active.

An AAA study in the summer found that losing driving privileges almost doubled the risk of increased depression symptoms in older adults.

It’s also important to note the huge challenge of predicting the full implications of a disruptive technology, such as autonomous driving. The same software developments could make drones or coolers on wheels capable of making many deliveries, which would have an impact on the number of trips being made. If better systems make mass transit more common — even if it’s just four neighbors riding together in a self-driving car to work — that would impact traffic, too.

As Silberg said, “We’re talking about the future, so we know it’s wrong.”