Robotaxis once lived only in science fiction. Now they pick up paying riders in a handful of American cities. Waymo went from the DARPA Grand Challenges to a commercial service in California in barely more than a decade, though the first cars still carried a safety driver. Backers of the technology, which has pulled in at least $100 billion, promised two things: safer streets and lighter traffic. The safety claim holds some weight. The traffic claim is where fresh data gets awkward.
Key Takeaways:
- A new MIT study found Waymo drives about 44 percent of its California miles with no passenger aboard — a pattern the industry calls deadheading.
- That empty-mileage rate roughly matches ride-hailing services like Uber and Lyft, which undercuts the pitch that robotaxis ease congestion.
- Buses, trains, and subways move far more people per foot of road, yet transit funding lags far behind the money pouring into robotaxis.
Waymo’s safety record stands up reasonably well. Company figures last year showed its cars crashed less often than human drivers and drew far lower insurance claims. The record isn’t spotless — recent trouble with school buses and flooded roads proved the cars still get caught out.
Traffic is the harder sell. Waymo’s own filings to the California Public Utilities Commission point the other way, and a new study reads them closely. Published in Transport Findings by Awad Abdelhalim, Assistant Director of Research at the MIT Transit Lab, it covers data from August 2023 through December 2025, roughly 1,000 days. Over that stretch, Waymo finished 13.8 million trips for 19.3 million passengers and covered 86.3 million miles (138.8 million km), with volume climbing about 15 percent each month. Abdelhalim chased one figure: how many of those miles rolled by with nobody in the back seat.
At the start, just 36 percent of Waymo’s miles carried a passenger. By the end of the study that share rose to roughly 56 percent, then flattened. So about 44 percent of the company’s miles happen with an empty electric car on the road. Anyone who has walked around San Francisco lately knows the sight: sensor-covered Jaguar I-Paces everywhere, most of them carrying no one.
Two kinds of empty
Deadheading splits into two types. One is a car circling while it waits for a ride request. The other is a car heading out to fetch a passenger. Waymo has trimmed the pickup miles as its fleet has grown, and the empty miles per trip keep falling — partly because the company added freeway routes, Abdelhalim suggests.
A separate analysis backs him up. Matthew Raifman, who studies policy and autonomous vehicles at UC Berkeley, ran the CPUC figures from January 2024 through September 2025 and landed on the same 44 percent. He also found that two-thirds of those empty miles were cars roaming while they waited for a customer.
That empty cruising isn’t just a line in a spreadsheet. As Waymo spreads into more cities, people have started to notice the cars looping past their homes with no one inside. Urban planners tracking the same CPUC data have flagged how much time the fleet spends repositioning rather than sitting still (Planetizen). In one Atlanta neighborhood, dozens of empty robotaxis funneled through a dead-end street before dawn until residents propped up a children’s traffic sign to confuse them. The denser the fleet, the more visible the cost of cars that never switch off.
We’ve heard this pitch before
The traffic promise sounds familiar because it is. In 2014, MIT researchers argued ride-hailing would shrink car ownership and ease congestion. Two of those authors later reversed course once the evidence showed ride-hailing added traffic and CO2, partly because cheap fares tempted people into trips they would not otherwise take. They warned robotaxis would likely fall into the same trap. A 2018 study pinned nearly half the rise in San Francisco’s vehicle miles on ride-hailing.
About 40 percent of the miles a Lyft or Uber driver logs are also empty. Put the two side by side and the congestion math barely budges whether a person or a computer steers. That same gap explains part of the robotaxi safety edge: a robotaxi usually carries fewer people than a staffed ride-hailing car, so its expected injury rate per mile comes out lower.
The bigger question is where the money goes. Robotaxis do carry a real safety record — Waymo’s insurance data points to far fewer claims than human drivers (NBC Bay Area). But moving people efficiently is a different problem, and the cheapest answer is old. One bus carries the load of dozens of cars; trains and subways do even better per foot of road. Transit just doesn’t draw the same checkbooks. Waymo raised $16 billion this year, and the sector has soaked up at least $100 billion since the 2010s. By contrast, the American Public Transportation Association asked for $268 billion over five years, and Transportation For America priced a top-tier transit network at $4.6 trillion across two decades.
For now, the robotaxis keep rolling — many of them empty — while the cheaper fix for congestion waits on funding that may never come.
Written by Vytautas Valinskas
