Fully self-driving cars are already on public roads, but only in a handful of cities and only as taxi services. For most people, the technology won’t feel “common” until the mid-2030s at the earliest, when self-driving features are expected to appear in a significant share of new cars sold and robotaxi services expand well beyond their current footprint.
Where Self-Driving Cars Operate Today
Waymo currently runs driverless robotaxis in several U.S. cities, including San Francisco, Phoenix, Los Angeles, and Austin. In late 2025, it announced expansion into five more cities: Miami, Dallas, Houston, San Antonio, and Orlando, with rides opening to the public in 2026. These vehicles operate without a human safety driver, navigating traffic, pedestrians, and intersections entirely on their own.
That said, these services come with limits. They work within carefully mapped zones in each city, typically covering core urban areas rather than entire metro regions. The cars avoid conditions they can’t yet handle reliably, like heavy snow or unmapped rural roads. This is what the industry calls Level 4 autonomy: the car drives itself completely, but only within approved conditions and locations. Level 5, where a car could drive anywhere a human could with no restrictions, doesn’t exist commercially yet and likely won’t for a long time.
What the Sales Projections Say
Goldman Sachs Research projects that by 2030, about 10% of global new car sales will be Level 3 vehicles. These are cars that can fully take over in specific situations, like highway driving in clear weather, letting you look away from the road entirely. Fully autonomous Level 4 vehicles are expected to make up roughly 2.5% of new sales by 2030, a smaller slice than previously forecast.
The picture changes significantly by 2040. In an optimistic scenario, cars with Level 3 autonomy or higher could account for about 60% of all new light vehicle sales globally. Even in a more conservative forecast, that number sits around 40%. Adoption rates will vary by region. China is expected to lead, with self-driving vehicles potentially reaching 90% of all car sales by 2040. Europe could hit nearly 80%, while the U.S. is projected at roughly 65%.
Keep in mind that new car sales don’t equal cars on the road. The average car stays in service for over 12 years. So even if half of all new cars sold in 2035 have advanced self-driving capabilities, it would take until the mid-2040s before a majority of vehicles on the road actually have them.
Why It’s Taking So Long
The core challenge is perception: teaching a car to “see” the world as well as a human does, in every possible condition. Rain, fog, snow, and glare all degrade the sensors that self-driving cars rely on. Cameras lose contrast in heavy rain. Lidar, the laser-based sensor that maps the 3D environment, can be scattered by snowflakes. Radar handles bad weather better but offers less detail. Fusing all these sensors together to maintain a reliable picture of the road in a downpour or blizzard remains one of the biggest unsolved engineering problems in the field.
Then there are edge cases: the countless rare scenarios that humans navigate on instinct but that a computer has never encountered. A traffic cop waving you through a red light. A child chasing a ball into the street from behind a parked truck. Construction zones with hand-drawn detour signs. Each of these situations requires the system to improvise, and current AI handles improvisation far less gracefully than routine driving.
Proving safety at scale is its own hurdle. Google’s early self-driving cars logged a police-reportable crash rate of about 2.2 per million miles in Mountain View, California, compared to 6.1 for human drivers in the same area. No fatalities occurred, versus a California average of roughly one death per 108 million miles for human drivers. But researchers have calculated that autonomous vehicles would need to be driven hundreds of millions, sometimes hundreds of billions, of miles to statistically prove they’re safer than humans across all conditions. That kind of validation takes time.
The Cost Problem Is Shrinking
One reason self-driving features were once confined to experimental vehicles is sheer hardware cost. Lidar sensors that cost tens of thousands of dollars a decade ago have dropped dramatically. In 2025, a high-resolution automotive-grade lidar unit runs between $600 and $1,500 for passenger cars with partial self-driving features. Entry-level systems for simpler driver-assistance functions cost as little as $150 to $300 per sensor. For fully autonomous robotaxis, which need more sensors and redundancy, the price per lidar unit sits between $1,500 and $6,000.
Solid-state lidar, a newer design with no moving parts, could cut these prices by up to 50% in the next few years as production scales. That matters because it brings the technology within reach for mid-priced consumer vehicles, not just luxury cars or commercial fleets.
Robotaxis Could Beat Private Ownership
One path to self-driving cars becoming “common” doesn’t require you to buy one at all. Robotaxi services are projected to get cheap enough that many urban drivers would save money by ditching their personal car. Today, a robotaxi ride costs roughly $8 per mile, far more than driving your own car. By 2035, McKinsey projects that cost could fall to about $1.32 per mile as sensor prices drop and fleet operations become more efficient.
At that price point, the math shifts. Currently, in a city like Washington, D.C., owning a car is cheaper than using ride-hailing services if you drive more than about 2,000 miles a year, which most people easily do. But with autonomous taxis, that breakeven point jumps to around 7,500 miles per year. For anyone driving less than that, calling a robotaxi for every trip would actually cost less than car payments, insurance, maintenance, and parking combined. Goldman Sachs projects a global fleet of a few million commercial autonomous vehicles in use for ridesharing by 2030.
Public Trust Is Still Low
Technology and cost aren’t the only barriers. Most people simply aren’t comfortable with the idea yet. A 2024 consumer readiness index from J.D. Power and MIT scored public acceptance of fully self-driving vehicles at just 39 out of 100. That’s a slight improvement after two years of declining trust, but it still signals deep hesitation.
The biggest concerns aren’t about crashes. About 64% of consumers worry that data collected by self-driving cars isn’t safe and secure. Over 80% want to understand what’s being done to prevent hacking, and 84% are concerned about data privacy overall. Most people aren’t confident their automaker is transparent about how driving data is being used. Until companies address these fears directly, adoption will lag behind what the technology can deliver.
A Realistic Timeline
If you’re wondering when you’ll regularly see self-driving cars in your daily life, here’s a rough breakdown. By the late 2020s, robotaxi services will likely operate in 15 to 20 major U.S. cities, but you’ll still notice them as a novelty. By the early 2030s, new cars with highway self-driving features (Level 3) will be common enough at dealerships that you’ll have the option when buying a mid-range vehicle. By 2035 to 2040, self-driving capabilities at Level 3 or higher could be standard in most new cars sold, and robotaxis could be a routine, affordable transportation option in most large cities.
The transition to a world where most cars on the road can drive themselves, not just most new cars being sold, likely won’t happen until the 2040s or even 2050. The technology is real and improving rapidly, but fleet turnover, regulatory frameworks, consumer trust, and the sheer difficulty of handling every possible driving scenario mean the shift will be gradual rather than sudden.