Anyone want to buy a car that drives itself?

Earlier this year, a mysterious new company called Tensor announced itself to the world by claiming it would be the first to sell fully autonomous vehicles to customers at scale. The news didn’t make much of a splash. No one had ever heard of Tensor, so it was the kind of announcement easy to dismiss as vaporware. But the idea was not unfamiliar. In fact, some of the world’s biggest companies are interested in selling driverless cars to individual customers. After all, if you can already hail a fully autonomous vehicle, why not own one too?

But it won’t be easy. The technological and legal hurdles are immense. Today’s robotaxis are restricted in where they can travel and under what conditions; it’s unclear whether people would accept similar limitations for a vehicle they owned. And mishaps, like driving through an active police scene as Waymo recently did, becomes even more alarming when it’s a privately owned vehicle.

But for the companies investing hundreds of billions of dollars into the technology, it was always inevitable that robotaxis would just be the beginning.

For years, experts dismissed the idea of privately owned autonomous vehicles, arguing that the technology was too expensive for private sales. Instead, fleet-owned robotaxis were the safer bet, helping amortize the costs of all the sensors and high-powered computing — said to be in the hundreds of thousands of dollars — needed to enable the cars to drive themselves. But now the costs for much of that equipment, including lidar, is coming down, resurfacing the idea that autonomous vehicles can be cheap enough to sell to regular people.

There is a growing list of companies who are champing at the bit to sell fully autonomous vehicles, including Waymo, Tesla, Lucid, and General Motors. Even little players like Tensor, which was spun out from former robotaxi operator AutoX in China, are trying to grab a piece of this undefined market.

It may eventually be possible, but would it be practical? An autonomous vehicle has different requirements than a conventional one, with a myriad of sensors that need to be cleaned and calibrated on an almost daily basis. Would the owner be up to the task?

Some of the companies are working on self-cleaning sensors. Waymo, for example, has tiny wipers that keep its lidar sensors clean of dirt and debris. Tesla has said it is working on robot vacuums to keep its robotaxi interiors tidy. But neither company has outlined how regular people will be expected to maintain their privately owned robot car — if it ever comes to that.

Phil Koopman, an autonomous vehicle expert at Carnegie Mellon University, theorized that privately owned autonomous cars may work similarly to fractional jet ownership, in which you can buy your own jet but still hand it over to a management company that handles maintenance and operational logistics. The idea is that you can “own” a driverless car while outsourcing virtually all responsibility to third-party specialists.

“All the sensors have to go through calibration, maybe every day,” Koopman said. “But these sensors are pretty exotic. I doubt that you can just run them for five years with no maintenance.”

Another consideration: Most autonomous vehicles are pretty ugly. They require multiple sensors to perceive the world around them, and to serve as backup in case any one fails. The result is a vehicle festooned with a variety of sensors, which can make a vehicle look unattractive, said Omer David Keilaf, CEO of lidar company Innoviz. And aesthetics play a crucial role in the kind of vehicle people tend to purchase.

“Ugly cars are not sold,” Keilaf said, “regardless of the features they provide.”

This has long been Elon Musk’s beef with lidar. He has argued that the laser sensors are too expensive and too obtrusive to work for personally owned vehicles. Tesla uses a camera-only system for its partially autonomous Level 2 products, including Full Self-Driving, that Musk has promised will eventually lead to fully autonomous vehicles. But that decision has also led to multiple delays in Tesla’s plans to sell autonomous vehicles to its customers.

This may be a temporary problem. Keilaf said that next-generation lidar is dramatically smaller and cheaper than earlier versions, which opens new opportunities for discreet mounting locations that don’t disrupt a vehicle’s appearance. Fewer sensors will also be key in making the car more palatable to regular people. Several sensor companies, including Innoviz, are working to reduce the number needed to enable higher levels of automation.

If you read between the lines, you can already see how some of the companies pitching this idea of privately owned autonomous vehicles are allowing themselves enough wiggle room, in case the technology doesn’t pan out.

For example, both Tensor and Lucid have said they would sell vehicles that are defined as Level 4, meaning they can operate without any human interventions, but only under specific conditions. That could mean certain geographic areas, like places without defined road markings and signs, could be off-limits. It could also mean the vehicle wouldn’t be able to handle certain weather conditions, like a downpour or a blizzard. In those instances, control may need to shift to the human driver.

It’s unclear how these limitations will be received by consumers. While Waymo remains popular in the handful of markets in which it operates, and Tesla still has its fans, surveys have found most people remain deeply skeptical of autonomous vehicles. Moreover, advanced driver assistance features are still niche and have yet to go mainstream. It may be that the auto industry is too far out ahead of its own customer base.

This pressure comes from both new US startups, like Waymo, and global competitors, particularly Chinese automakers that are deeply invested in AI-driven vehicles, said Steve Man, global lead director of auto and industrial market research at Bloomberg Intelligence. Automakers typically frame autonomy not only as a commercial competition but as part of a broader geopolitical contest over AI leadership, Man said. They feel like they must invest in autonomy — even without certainty about consumer demand — because being caught behind the curve could eventually put them out of business.

Whether people actually want to own a self-driving car seems almost beside the point. After all, car companies have been indulging the idea of privately owned autonomous vehicles for over 50 years. Take this video from 1956, produced by GM, depicting a family of four cruising along in a self-driving, bubble-topped sedan in the wondrous, far-off future of 1976.

“Ah, this is the life,” the father says after stowing the steering wheel away. “Safe, cool, comfortable.”

This is an idea that has long permeated the mythmaking of the American auto industry, and likely will continue, even if the promise turns out to be more of a fantasy.