Thursday, November 9, 2023

Robotaxi Economics

The New York Times team of Tripp Mickle, Cade Metz and Yiwen Lu have been covering San Francisco's experiment with robotaxis from Waymo and Cruise for some time. Their latest report is G.M.’s Cruise Moved Fast in the Driverless Race. It Got Ugly., and it is a doozy. Leaving aside the serious safety issues that finally forced the DMV to stop Cruise operating, their history of interference with first responders, and the traffic congestion they exacerbate, the really devastating part of the article is the economics. Below the fold, I lay out the numbers.

The reason robotaxis look like a good investment is that the up-front investment in equipping cars with high-tech sensors and powerful computers would earn returns in two ways, by removing the need to pay drivers, and by operating 24/7. In the real world, it isn't quite that simple.

First, the investment is very significant. The NYT team reports on Waymo:
The cars are staggeringly expensive, outfitted with high-tech sensors and cameras, and are worth as much as $200,000.
And on Cruise:
Each Chevrolet Bolt that Cruise operates costs $150,000 to $200,000, according to a person familiar with its operations.
So the upfront investment is around $150K/car. Assuming a 5-year life and 7% interest, the annual cost would be around $32K, or about $90/day. After paying about $12/hr for fuel, insurance, repairs, depreciation and so on an Uber driver makes $18/hr, or assuming an 8hr day, $144/day. The robotaxi looks like it is about $54/day ahead.

Proponents will claim that the robotaxi works 24hrs/day, not just 8. But the demand for rides isn't constant through the day. The robotaxi fleet needs enough cars to satisfy peak demand, so most will be idle for most of the time, still consuming interest and depreciation. Uber doesn't pay for the cars, and Uber drivers only work when there is enough demand for rides. I think this is roughly a wash, so the robotaxi stays about $54/day ahead.

So far I have assumed that the robotaxi fleet needs no humans. But, as the NYT team reports, that isn't the case:
Half of Cruise’s 400 cars were in San Francisco when the driverless operations were stopped. Those vehicles were supported by a vast operations staff, with 1.5 workers per vehicle.
With 1.5 workers at Uber's pay rate the robotaxi's staff and equipment is costing $306/day, so the robotaxi is $162/day behind, and Cruise is unlikely to be able to pay only $18/hr. What were the 1.5 people per car actually doing?
The workers intervened to assist the company’s vehicles every 2.5 to five miles, according to two people familiar with is operations.
Lora Kolodny followed up the NYT team's reporting in Cruise confirms robotaxis rely on human assistance every four to five miles:
Vogt confirmed that the General Motors-owned company does have a remote assistance team, in response to a discussion under the header, "GM's Cruise alleged to rely on human operators to achieve 'autonomous' driving."

The CEO wrote, "Cruise AVs are being remotely assisted (RA) 2-4% of the time on average, in complex urban environments. This is low enough already that there isn't a huge cost benefit to optimizing much further, especially given how useful it is to have humans review things in certain situations."
Clearly, the robotaxi fleet needs some staff. The cars need refueling, say every 200 miles. They need cleaning. Lets guess the average ride is 5 miles and they need cleaning every 10 rides. That would be staff required every 40 miles. So the vast majority of interventions are because Cruise's self-driving technology just can't handle San Francisco traffic. They may be called robotaxis, just pay no attention to the driver behind the curtain. These remote drivers will be more expensive than the in-car drivers they are replacing!

Overall, their robotaxi operation is costing GM a boatload of money:
G.M. has spent an average of $588 million a quarter on Cruise over the past year, a 42 percent increase from a year ago.
$2,232B/year for 400 cars is $5.6M per car per year! How could each car earn this much?
The company charged an average of $10.50 per ride in the city.
To generate $5.6M each car would have to service 510K rides/year or 1,400 rides/day, or one ride per minute. Cruise's goal, before they got shut down, was less ambitious:
The shutdown complicates Cruise’s ambition of hitting its goal of $1 billion of revenue in 2025.
Covering less than half their costs in 2025 is about 91M rides, or about 250K/day. Lets be generous, and assume 10 rides/day/car. They would need 25K Chevrolet Bolts. Scaling the operation from 400 to 25K in under 2 years would be hard. It would be made harder because Bolt production stops next month. But even if the Bolts were available, this would be an additional investment of between $3.7B and $5B in order to lose only $1.2B/year.

I'm not the only one pointing out that the numbers don't add up. Here, for example, is Atrios:
Focusing on the notoriously high margin taxi business has always been odd, in that even if they "work" it's not entirely clear where the big profit opportunity is! Replacing low wage workers with expensive machines that require a lot of attention from people who have more specialized skills than "driving" doesn't seem to be such a great plan!

I get the appeal for the carbrained. What "we" all want is a 24 hour driver at an affordable rate. It's just not quite clear what the business model is, even if the technology works!

People always point to niche cases in which robot car would be particularly useful. But they're niche cases and you need scale to make anything in this industry profitable! Scale is necessary for a lot of reasons, but one is maintenance and spare parts production. Your self-driving campus vans won't last long if there's only one guy who knows to maintain them and the parts manufacturer just went bankrupt.
Last August, Hubert Horan continued his long-running series with Part Thirty-Three: Uber Isn’t Really Profitable Yet But is Getting Closer; The Antitrust Case Against Uber, which starts:
After $33 Billion in Losses Over 14 Years, Uber is Finally Approaching GAAP Breakeven

Uber claimed its first ever quarterly GAAP profit when it released its second quarter and first half financial results on August 1. The claim was a bit of stretch as the reported $394 million second quarter profit ($237 million for the first half) was entirely explained by an alleged $386 million second quarter gain ($707 million in the first half) in the value of untradable securities they hold in companies like Didi, Grab, and Aurora that have nothing to do with their ongoing operations. Readers of this series will know that Uber has aggressively used claims like this to justify misleading claims about its corporate financial performance ever since 2018 when it inflated published net income numbers by $5.8 billion just prior to its IPO.
Uber has always viewed drivers as a temporary necessity to tide them over until robotaxis take over and boost them into actual instead of manipulated profits. Last month they started offering robotaxi service in Phoenix using Waymo's cars. But unless Waymo is showing them radically different numbers than the NYT team found for Cruise, Horan's series is going to run and run.

These numbers look even worse for Tesla. Last year Matthew Loh reported that Elon Musk says the difference between Tesla being 'worth a lot of money or worth basically zero' all comes down to solving self-driving technology, and the reason was that owners would rent out their Teslas as robotaxis when they weren't using them. This was always obviously a stupid idea; who wants drunkards home-bound from the pub throwing up on their Tesla's seats? But the fact that the numbers don't add up for robotaxis in general, and the fact that Hertz is scaling back its EV ambitions because its Teslas keep getting damaged because half of them are being used by Uber drivers as taxis, make the idea even more laughable.

It is really surprising that the DMV took so long to shut Cruise down. In August the NYT team wrote:
Last week, a Cruise driverless car collided with a fire truck responding to an emergency. Another Cruise vehicle got stuck in wet concrete. The week before, several Cruise cars blocked traffic in the city’s North Beach neighborhood.
More recently, Sam Biddle reported that:
previously unreported internal materials such as chat logs show Cruise has known internally about two pressing safety issues: Driverless Cruise cars struggled to detect large holes in the road and have so much trouble recognizing children in certain scenarios that they risked hitting them. Yet, until it came under fire this month, Cruise kept its fleet of driverless taxis active, maintaining its regular reassurances of superhuman safety.
In contrast, all the NYT team could find to report about Waymo was:
In May, one of its cars struck and killed a small dog. A few years ago, a driverless Waymo car with a human safety driver operating the wheel hit a pedestrian who needed to be taken to the hospital. The company has been collecting fares in the Phoenix area for several years and now has a fleet navigating some 200 miles across that region, including to and from the airport.
Why is Cruise so much less safe? The NYT team write:
Company insiders are putting the blame for what went wrong on a tech industry culture — led by the 38-year-old Mr. Vogt — that put a priority on the speed of the program over safety. In the competition between Cruise and its top driverless car rival, Waymo, Mr. Vogt wanted to dominate in the same way Uber dominated its smaller ride-hailing competitor, Lyft.

“Kyle is a guy who is willing to take risks, and he is willing to move quickly. He is very Silicon Valley,” said Matthew Wansley, a professor at the Cardozo School of Law in New York who specializes in emerging automotive technologies. “That both explains the success of Cruise and its mistakes.”
As someone who has been sharing the roads with Waymo cars for many years, I have always believed that Waymo, unlike Cruise and much worse Tesla, understood that it was necessary to prove that self-driving technology was safe, not just gaslight the public as Mr. Vogt did:
To make streets safer, he said in an interview, cities should embrace self-driving cars like those designed by Cruise, a subsidiary of General Motors. They do not get distracted, drowsy or drunk, he said, and being programmed to put safety first meant they could substantially reduce car-related fatalities.
Two months later Cruise's safety record got them shut down. The very same day:
CEO Mary Barra told financial analysts that Cruise “is safer than a human driver and is constantly improving and getting better.”


dr2chase said...

My brief encounters with Waymo (then, Google) cars left me with a similar impression (note a conflict of interest, I work for Google). One was watching a car take a turn off Alma west onto Rengstorff -- the car paused before crossing the tracks. The other was on a bicycle, traveling behind the Google campus across the entrance to some park (Shoreline?). I was on-road for a bit, I heard a car behind me, and I did all the things that years of biking in the company of human drivers have trained me to do, expecting the sort of "I am important and in a hurry" pass that humans always do. But the car did not pass, did not even crowd, in a way that was so unusual that I looked, and the inhumanly polite driver, was in fact, not a human.

Even assuming inhuman politeness, I am not sure that having these on the road would be a net-positive for years; the cost seems so high that their initial use would be for wealthy long-distance commuters who want to work/sleep/whatever during their commute. Their use as cabs, assuming the numbers ever work, would probably lead to an uptick in curb pickups and dropoffs, which creates conflicts with people on bikes (depending on infrastructure) and convenient car storage (seems boneheaded to reserve 1/3 of a road or more for 0-mph 0-occupancy automobiles, but we do). That's a problem now with human Lyft and Uber drivers; at least some of them are notably terrible.

David. said...

Richard Lawler reports that:

"Cruise robotaxi shutdown expands, pressing pause on supervised and manual trips too.

Cruise’s robotaxis are front and center of the industry’s trust issue after losing their California permit following an incident where a pedestrian ended up stuck underneath one of its cars.

It already halted service nationwide and said it’s installing new updates. Now Cruise has announced it’s taking its cars off of public roads while it undergoes a full safety review."

David. said...

Kirsten Korosec and Rebecca Bellan report that Cruise co-founder and CEO Kyle Vogt resigns. Moving fast and breaking things may not have been the best approach to autonomous vehicles.

David. said...

Brandon Vigliarolo reports that California commission says Cruise withheld data about parking atop of a pedestrian:

"California's Public Utilities Commission (CPUC) has threatened driverless taxi outfit Cruise with fines and sanctions unless it can prove it didn't withhold information and make misleading statements about an October accident.

According [PDF] to CPUC administrative law judge Robert Mason, Cruise officials neglected to tell the CPUC that a pedestrian struck by a Cruise vehicle on October 2 was dragged 20 feet before the car parked on top of her. Cruise has been in a tailspin since the incident, with a national pause on operations and a reshuffle in its management.

"[Cruise spokesperson Jose] Alvarado's description of the incident only included that the Cruise AV immediately stopped upon impact with the pedestrian and contacted Cruise's remote assistance," Mason said.

"Mr. Alvarado's description of the October 2, 2023 incident omitted that the Cruise AV had engaged in the pullover maneuver which resulted in the pedestrian being dragged an additional 20 feet at 7 mph," the judge added, citing testimony from the CPUC analyst first informed of the incident."

David. said...

In 7.1 million miles, 3 minor injuries: Waymo’s safety data looks good Timothy B. Lee reports that:

"Waymo on Wednesday released new crash data based on the company's first 7.1 million miles of fully driverless operations in Arizona and California. The data shows that human-driven cars are more than twice as likely to get into a crash that is reported to the police. And depending on how you do the math, human-driven cars are four to seven times more likely to get into crashes that lead to an injury.

Through October 2023, driverless Waymo vehicles have had only three crashes with injuries—two in the Phoenix area and one in San Francisco. Waymo says all three injuries were minor. If those same miles had been driven by typical human drivers in the same cities, we would have expected around 13 injury crashes.
After making certain adjustments—including the fact that driverless Waymo vehicles do not travel on freeways—Waymo calculates that comparable human drivers reported 1.29 injury crashes per million miles in Phoenix and 3.79 injury crashes per million miles in San Francisco. In other words, human drivers get into injury crashes three times as often as Waymo in the Phoenix area and six times as often in San Francisco.

Waymo argues that these figures actually understate the gap because human drivers don't report all crashes. Independent studies have estimated that about a third of injury crashes go unreported. After adjusting for these and other reporting biases, Waymo estimates that human-driven vehicles actually get into five times as many injury crashes in Phoenix and nine times as many in San Francisco."

Tardigrade said...

An important question on the Waymo crash data is how many human-caused crashes are due to bad maintenance (such as insufficient tread on tires). Presumably the Waymo cars are all properly maintained. Still, these numbers do look very good so far.

David. said...

There is too much good stuff on Cruise's disaster in San Francisco and their competition with Waymo to excerpt from Brad Templeton's Robocar 2023 In Review: The Fall Of Cruise. Just go read the whole ting.

David. said...

GM's driverless car company Cruise is under investigation by several agencies by Camila Domonoske reveals that:

"The company on Thursday said it is being investigated by the National Highway Traffic Safety Administration, the U.S. Department of Justice, and the U.S. Securities and Exchange Commission, in addition to California agencies. Cruise said it is "fully cooperating" with the regulatory and enforcement agencies that have opened the investigations."

David. said...

Aarian Marshall confirms that Cruise failed to disclose disturbing details of self-driving car crash:

"The new report by law firm Quinn Emanuel says that Cruise failed to tell California’s Department of Motor Vehicles that after striking a pedestrian knocked into its path by a human-driven vehicle, the autonomous car pulled out of traffic—dragging her some 20 feet. Cruise said it had accepted the firm’s version of events, as well as its recommendations.

The investigators found that when Cruise played a video of the crash taken from its autonomous vehicle for government officials, it did not “verbally point out” the vehicle’s pullover maneuver. Internet connectivity issues that occurred when the company tried to share video of the incident “likely precluded or hampered” regulators from seeing the full video, the report concluded.
Another third-party report on the crash released by Cruise today, by the engineering consulting firm Exponent, found that technical issues contributed to the autonomous vehicle’s dangerous pullover maneuver. Although the self-driving car’s software correctly detected, perceived, and tracked the pedestrian and the human-driven car, it classified the crash as a side-impact collision, which led it to pull over and drag the woman underneath it."