Investors have poured billions into startups promising to get self-driving cars on the road to shuttle passengers and parcels. Now, the startup world has largely shifted to applying autonomous driving to specific industrial use cases — in an effort to build a business much faster.

Why it matters: Despite bold predictions that we'd soon be riding in robotaxis around town, the task is taking much longer than expected and is a lot more expensive.

• While established players will surely continue to find capital to fund their work, the math has become much less attractive to newcomers, especially as investors look for companies with clearer business models.

The big picture: The market for companies building fully self-driving cars that can give passengers rides has greatly consolidated, with a few large, well-funded players still pursuing that goal.

• And while they’ve made inroads and deployed pilots in cities like San Francisco and Phoenix, they’re nowhere close to operating a self-driving Uber-like service.
• Not even Uber is still trying to be Uber in this space. Uber and Lyft have sold off their autonomous driving divisions after sinking billions of dollars into them. Apple, meanwhile, has been floundering with its self-driving car ambitions.
• Others like GM-owned Cruise have started to test delivery capabilities in addition to its ride-hailing aspirations.
• Even Aurora, one of the early hopefuls of this category — and reportedly in talks to merge with a SPAC — has shifted its focus to trucking.

Between the lines: "Robotaxi, for a while, was the Auto-Tune of the autonomous vehicle world," Trucks Venture Capital managing partner Reilly Brennan explains. It was the shiny and attention-grabbing focus of the industry for a while, but that’s changed.

• "We have not seen a lot of robotaxi ideas from the early-stage in years," he says, contrasting it with 2015 through 2017, when his firm would see new startup ideas in this segment almost weekly.

State of play: Now, startups are focused on applying autonomous driving to more specific use cases and industries like agriculture and mining.

• This means automated driving that is low-speed, off public roads, not carrying passengers and has a high business value to customers.
• "If you're doing a Series A company right now and you're doing robotics and automation, you need to have revenue," says Brennan.
• A lot of the companies are also building software and other tools for for advanced driver-assistance systems (ADAS), coming to market more quickly.

The upside: These industry-focused companies could become much more valuable than what we've seen so far from self-driving car startups, thanks to their ability to get to market much faster and the potential to build large businesses if they truly provide more efficiency to their customers.

• For context: Together, robotaxi companies Cruise, Zoox and nuTonomy were acquired for about $2.3 billion.

The bottom line: "My personal opinion about this is that we're never gonna have Level 5," predicts Brennan, referencing the classification for fully self-driving cars.

• "Only humans are dumb enough to think they can drive in all conditions — robots are smarter than that."

What’s next: "If 2020 and early 2021 was defined by electric vehicle SPACs, it appears that the rest of this year is gonna be autonomous SPACs," predicts Brennan, who notes a change in the types of companies sponsors have been recently asking him about.

It’s been more than a decade since Google began working on its self-driving car project, yet the U.S. is still kicking the tires on comprehensive regulations for autonomous vehicles.

Why it matters: A number of companies have been testing their autonomous vehicles for several years now, but the lack of inroads in regulations will keep those efforts in test mode.

Driving the news: In April, Sens. Gary Peters (D-Mich.) and John Thune (R-S.D.) introduced an amendment — which ultimately didn’t make the cut for the U.S. Innovation and Competition Act (formerly, Endless Frontier Act) — that would have given the National Highway Traffic Safety Administration (NHTSA) the ability to exempt 15,000 vehicles per manufacturer from certain safety standards.

• The bill would have given more leeway for companies to produce and deploy cars not designed for human drivers.
• Right now, NHTSA can only exempt up to 2,500 vehicles per manufacturer, and has only handed a waiver to one company: Nuro, in early 2020.

The big picture: Currently, vehicles are regulated by the federal government, while "drivers" are overseen by the states.

• "One of the big issues with AVs is that this very normal distinction [between vehicle and driver] for regular cars collapses," explains Ed Niedermeyer, communications director at Partners for Automated Vehicle Education (PAVE).
• "This is one of the first areas where we have to make some decisions about the legal personhood — or the extent of the legal personhood —of AI." This means tackling complex questions regarding legal liability and being able to measure self-driving software's "safety."
• In November, NHTSA put out a call for public comments on how it should structure a safety framework for "automated driving systems" (that is, autonomous vehicles).
• The main questions raised by commenters include the agency's regulatory scope, safety standards, addressing car-to-car connectivity, third-party verification and safety for humans outside the vehicle.

Meanwhile: There's a wide range when it comes to how individual states regulate autonomous driving.

• In California, dozens of companies have obtained permits to test autonomous driving cars with safety drivers (and eight are also licensed to do so without drivers). The state has a number of requirements, including that they report data on their miles driven and rates of disengagement (that is, how often a human driver has to take over to maintain safe driving).
• In Florida, where a few companies like Ford's Argo AI have rolled out pilots, there are almost no restrictions as long as companies meet safety and insurance requirements and report accidents.

The bottom line: There’s still a long road ahead.

Futuristic auto tech, from electric vehicles to sensors used by self-driving cars, was a magnet for special purpose acquisition companies over the past year, but those companies haven’t fared so well in the public markets.

Why it matters: If SPAC sponsors are going to continue pursuing these companies, they'll have to pick ones that are more ready to meet investor expectations of hitting milestones and growing their businesses, or more heartbreak will be in the cards for everyone.

Catch up quick:

• Regulators are scrutinizing a number of companies, including Canoo and Lordstown Motors, with the latter telling investors this week that it may run out of cash before it can begin production of its cars later this week.
• Fisker, Canoo, Lordstown and Nikola are trading between 72% and 90% lower than their highs, and shares of 19 EV and EV-adjacent companies that have gone public via SPACs have seen their share prices fall about 50% from their highs, as Axios recently reported.

• Electric vehicle SPACs lose their luster (Axios)
• AI pioneer Raquel Urtasun launches self-driving technology startup (TechCrunch)
• Self-driving cars speed forward while safety rules lag (Axios)
• Self-Driving Cars Could Be Decades Away, No Matter What Elon Musk Said (WSJ)

Anthony Levandowski, the engineer and entrepreneur at the center of a yearlong legal battler between Waymo and Uber (and who was later pardoned from his criminal conviction of stealing trade secrets), has been working on autonomous driving long before joining either company.

• Question: What vehicle did he and his team build for a DARPA Grand Challenge that was eventually donated to the Smithsonian? (Answer at the bottom.)