After years of positive vibes about the future of autonomous vehicles and nearly unrestricted access to cash from Kool-Aid-drunk venture capitalists, the AV industry is confronting some hard truths. The first is that autonomous vehicles are going to take a lot longer to reach mass scale than previously thought. The second is that it’s going to be a lot more expensive, too. And the third hard truth: going it alone is no longer a viable option.

Last week, Lyft sold its self-driving car division to a subsidiary of Toyota for $550 million. Cruise bought Voyage. Aurora merged with Uber’s autonomous vehicle unit. Delivery robot startup Nuro acquired self-driving truck outfit Ike. There have been so many mergers, joint ventures, and various tie-ups lately it can be difficult to keep them all straight.

Where that leaves things is a little unclear. There is still money flowing to these companies, and nearly all of the executives, engineers, and software developers working on the technology remain bullish about the future. But there is a growing sense among experts and investors that the heady days when anyone with a couple of test vehicles, some LIDAR, and a vision for the future could launch a startup are at an end. And there will definitely be more shrinkage to come.

“The consolidation is long overdue,” said Raj Rajkumar, robotics professor at Carnegie Mellon University. “The AV industry has promised too much for too long, and has delivered too little.”

There are other signs of general unease in the AV world. Last month, John Krafcik announced that he was stepping down as CEO of Waymo after helping lead the company since 2015. Krafcik oversaw the transformation of Google’s self-driving car division from “Project Chauffeur” to its own standalone company. During his time, Waymo struck partnerships with several automakers, launched a limited ride-hailing pilot with fully driverless vehicles in Arizona, and expanded its commercial ambitions to include trucking and last-mile delivery.

As a former auto executive with years of experience at Hyundai and Ford, Krafcik was seen as someone who could bring self-driving cars to the mainstream. But his decision to step aside at a time of heightened uncertainty highlights the difficult road ahead for self-driving cars — especially as many of those early predictions have failed to come to pass.

Krafcik was not able to bring to market an autonomous vehicle that can drive on any road and in any conditions without years of rigorous testing. Aside from a small area outside Phoenix, Waymo’s cars require backup drivers to monitor the vehicle’s operations and take control if anything goes wrong.

Waymo still has a commanding lead in autonomous vehicles, but diminished expectations about the future of self-driving cars are affecting its business in other ways, too. In 2018, Morgan Stanley published a research note valuing the company at $175 billion based on its perceived leadership in the space and the potential for massive amounts of revenue from its autonomous vehicles.

A year later, the bank slashed Waymo’s valuation to $105 billion, still an extraordinary amount of money for a company with no revenue. But last year, when Waymo announced its first external funding round of $2.25 billion, the company declined to discuss what valuation that investment was based on. Later, the Financial Times reported it to be $30 billion — a nearly 85 percent decrease from 2018.

Valuations aren’t the sole indicator of where things stand with autonomous vehicles. But they do point to investor sentiment. And right now, investors are not feeling as confident in the prediction that autonomous vehicles will soon become the dominant form of transportation.

For years, Missy Cummings, director of the Humans and Autonomy Lab at Duke University, has been criticizing rosy predictions about our driverless future. She’s consistently warned that the technology is much further away and harder to get right than anyone in the industry cares to admit.

The recent trend in consolidation is vindication for her position, she says.

“It’s kind of like the elephant in the room,” she said of the shrinking of the AV world. “People will mention that and then they’ll stop themselves from making the Socratic connection to what this means about the viability of this industry.”

But Cummings doesn’t think people in the industry will be able to ignore the truth for much longer. “There is an embarrassingly large sum of money that’s been invested in this, so people feel like they have to keep going down that path because surely all these people who invested all this money can’t be wrong,” she says.

“Not everyone is delusional,” she added. “Just most people in this business.”

That said, Toyota and Aurora weren’t delusional when they decided to buy the automated driving teams at Lyft and Uber, respectively. They likely saw the value in the code produced by those teams, as well as the talent accrued by the ride-hailing companies over the years. When you can’t hire the people you’d like to staff your own projects, then you have to acquihire them, the distinctive Silicon Valley practice of buying a smaller company for the express purpose of acquiring their team of software engineers. Also, Uber and Lyft were very motivated to sell as recently public companies under pressure to staunch the bleeding and become profitable.

Another recent acquihire was when Apple bought bankrupt self-driving startup Drive.ai back in 2019. Apple had just laid off over 200 engineers from its secretive Project Titan, its own floundering self-driving operation. The Drive.ai team, composed of several graduates from Stanford’s AI lab, had already had some modest success with a ride-hailing pilot in Texas. Unhappy with its own team, Apple simply fired them and bought a new one.

“The buying up of these companies represents companies being able to buy skill sets that they would not otherwise be able to recruit,” Cummings said. “And I think that’s very valuable.”

The recent reorganization of the self-driving car industry may also point to a new set of priorities. Namely, robotaxis are out, and logistics and industrial applications are in.

Four years ago, the industry saw plenty of startups come out of stealth promising to do everything: they were going to make hardware and software; they were going to build their own vehicle; they were going to do robotaxis and delivery and trucking. And of course, they were going to change the world in the process. “It was a little bit like the decathlete business model,” said Reilly Brennan, general partner at venture capital firm Trucks. “We’re going to be very good at 10 different things.”

Now, most investors are interested in more “structured” applications of automated driving technology, such as construction, mining, middle-mile delivery, and agriculture. “So instead of being a decathlete, you had to pick a javelin thrower,” he said.

That’s not to say robotaxis are completely dead — far from it, with major players like Waymo, Cruise, Argo, and Baidu as well as smaller startups like Pony.ai, May Mobility, and Optimus Ride still banking on the proliferation of autonomous ride-hailing vehicles by the middle of this decade. But Brennan says the robotaxi craze has definitely cooled in recent years.

“Frankly, we stopped seeing robotaxi startups in probably the end of 2017,” Brennan said. “Here we are in 2021 and there are very few ongoing robotaxi startups that aren’t, you know, Cruise, Waymo, or Argo.”

Brennan thinks the concept of autonomous vehicles that can drive anywhere under any conditions is still possible, but much further out than originally thought. They will require huge financial investments — “11 figures” by Brennan’s estimates — and a willingness to tolerate zero cash flow until the technology is mature and safe enough to launch. There are only so many companies with deep enough pockets to take on that challenge: major car companies like Ford, GM, and Volkswagen or tech giants like Apple, Alphabet, and Intel. Everyone else is probably not long for this world.

The mid-level engineers always knew this to be true, Brennan said. It was the CEOs who were making the erroneous predictions about the availability of self-driving taxis by 2020. “I think the CEOs of those companies knew that they were going to be playing golf by 2020,” he said.