Aurora’s stock had a wild ride this month and CEO Chris Urmson, who led Google’s self-driving car program in its pre-Waymo days, thinks it’s on the cusp of commercializing driverless trucks. But profitability is years away.
By Alan Ohnsman, Forbes Staff
The self-driving vehicle landscape is littered with the corpses of those who tried and failed. Last year saw General Motors pull the plug on Cruise; rivals Ford and Volkswagen scuttled their lavishly funded Argo AI venture in 2022 and Uber offloaded its self-driving tech unit in 2020 after a fatal accident. Then there are the smaller companies who died before many had even heard of them: TuSimple, Embark, Ike and Starsky Robotics.
Alphabet’s Waymo, the oldest and best-funded self-driving firm, is the big exception here, with robotaxi fleets in Phoenix, San Francisco and Los Angeles, and plans to expand into Austin, Atlanta and Miami. Forbes estimates its revenue topped $100 million last year and may rise substantially in 2025.
Then there’s Aurora Innovation. Founded in 2017 by three autonomous driving experts from Google, Tesla and Uber, it’s managed to stay alive and is the only pure publicly traded U.S. autonomous vehicle stock left. But more than three years after going public it hasn’t yet booked revenue. And it’s been forced to delay the launch of its driverless truck services — twice. Cofounder and CEO Chris Urmson says that won’t happen again, and come spring Aurora will get on a similar road to commercial success as its Alphabet-backed rival, which he led when it was incubating inside Google.
“We expected that this was a hard enough problem that there was only going to be a handful of people to solve it. … And it’s kind of played out that way,” Urmson told Forbes. “Waymo seems to have cracked it and we’re on the cusp of launching in April.”
That launch, which was first supposed to happen in 2023, will begin with a single semi-truck hauling loads between Dallas and Houston. Others will follow and Aurora hopes to have “tens” of trucks on the road by the end of the year, some of them traveling a second route between El Paso and Fort Worth.
It’s a purposely slow start. And even if all goes well, investors will need to be patient, likely for years, before Aurora generates revenue and profits to justify its $11.7 billion market cap. Aurora isn’t facing any immediate danger from the funding (it has more than $1 billion), safety or management problems that killed some competitors. But the timeline for when and if it enters the black remains fuzzy–assuming all goes well in April.
Meanwhile, the challenges, technical and business, are daunting: developing computing systems better at instantaneously processing visual data, finding cheaper and more powerful sensors for detailed images of road conditions and hazards, and in the case of trucking, building and wrangling a mess of relationships with freight companies and truckmakers.
“Transportation is a huge global industry, worth trillions of dollars, which is exciting, but it’s also generally low margin,” said Alysin Malek, who runs Middle Third, a mobility tech consultancy, cofounded autonomous shuttle startup May Mobility and was part of the GM Ventures team that acquired Cruise in 2016. “And you have to have a full suite of relationships–which is a lot of work to set up in trucking because it’s a highly fragmented industry.”
Investors are eyeing Aurora with cautious optimism. Early this month the company’s languishing shares, which traded below $5 for much of 2024, spiked 29% to close at $8.39 on news of tech partnerships with AI chip behemoth Nvidia and truck parts maker Continental. Since then, however, much of that gain has been lost. Aurora closed at $6.80 on Jan. 17. (Even at that price, the 146.3 million shares Urmson holds put his net worth around $1 billion.)
Aurora and remaining competitors like Kodiak and Waabi are developing AI-enabled big rigs for a simple reason: the trucking industry generates about $1 trillion of annual revenue and faces a shortage of human drivers. If they can successfully commercialize the tech, the autonomous truck business could be worth $600 billion annually by the mid-2030s, according to a McKinsey study.
Five or six years ago, operating autonomous semis on highways looked a lot easier than deploying robotaxis in big cities. That’s because highways are simpler than urban streets and have fewer obstacles to navigate. But highways demand higher speeds and the massive size of 18-wheelers, weighing up 80,000 pounds with the cab and a loaded trailer, present their own challenges. Throw in high winds, torrential rain or slick, icy conditions and the potential for terrible accidents increases. Even robotaxis can’t yet safely operate at highway speeds; Waymo cars drive at under 45 miles per hour.
George Mason University professor Missy Cummings, an artificial intelligence expert who advised NHTSA on autonomous vehicles, told Forbes she’s dubious that the highway problem has truly been solved. “I have yet to see ANY self-driving company–car or truck–reliably operate at highway speeds,” she said. “Even Waymo shut their autonomous truck program down quietly, but no one really talks about this.”
Robotic Revenue
Urmson, with decades of experience with AVs, says ensuring safety is one reason it’s taken longer than anticipated to shift to fully driverless operations on highways. Equipping Aurora’s trucks with powerful vision systems that have longer range than robotaxis is part of that.
“The ability to see as far down the road as you need to for trucking, I think we didn’t understand that back when I was at Google,” he said. “We started to get a handle on it but we didn’t know how to solve it.”
He thinks Aurora will do that with Firstlight, its proprietary laser lidar tech that generates 3D maps of distant road conditions to bolster camera and radar data.
Autonomous vision systems remain a persistent challenge, said Cummings. “The bottom line is that perception systems are quite brittle and they cannot be easily backfilled by remote humans. Unless one of the companies has some kind of tech breakthrough, I do not see highway operations occurring anytime soon.”
During its current pilot phase, Aurora is hauling loads with companies including FedEx, Uber Freight, Schneider, Werner and Hirschbach. When its driverless service launches, it expects to work with Uber Freight, Schneider and Hirschbach. By 2025’s second quarter, the company hopes to finally start reporting revenue from robotic services. It’s currently being paid by customers for hauling loads on training runs with safety drivers at the wheel). But Aurora’s not ready to set revenue targets or a date for when it will be profitable.
In the company’s investor letter last October, it reported hauling more than 8,200 commercial loads since September 2021 and racking up over 2.2 million test miles while doing so. The same release said it’s on a “path” to gross profitability by the end of 2026.
Getting close to the launch of its commercial product hasn’t been easy or cheap. Company filings show that it’s poured more than $2.4 billion into R&D from 2020 through last year’s third quarter. In that same period, cumulative losses totaled $3.7 billion.
Urmson’s not Auora’s only technical expert. His cofounders are Chief Product Officer Sterling Anderson, who helped create Tesla’s Autopilot, and Chief Scientist Drew Bagnell, a Carnegie Mellon University professor who was previously part of Uber’s autonomous tech unit (which Aurora acquired in 2020). Getting close to the launch of its commercial product hasn’t been easy or cheap. Company filings show that it’s poured more than $2.4 billion into R&D from 2020 through last year’s third quarter. In that same period, cumulative losses totaled $3.7 billion.
An Aurora filing this month indicates it may raise new funds at some point, but isn’t doing so now. It reported $263 million of cash at the end of 2024’s third quarter and with its short-term investments the company has more than $1.2 billion of total assets.
“The good news is we have a lot of cash,” Urmson said. “We feel like we’re in a very strong position.”
Meanwhile, competition in the robotic trucking space is starting to heat up. Canada’s Waabi is also working toward the start of its commercial operations, as is Bot Auto, a startup created by TuSimple cofounder Xiaodi Hou. Kodiak, led by a former Google self-driving car project computer scientist, began hauling sand off-road to oil and gas fields with driverless semis in West Texas last year, targeting a somewhat easier niche market. Silicon Valley’s Gatik, another startup, is avoiding highway trucking routes for now and hauling groceries, office supplies and other goods in smaller commercial trucks on city and suburban streets.
Shifting to commercial operations has taken longer than planned, but Urmson thinks Aurora is going to be the first company to do that in trucking.
“The work we’ve put into simulation, our approach to artificial intelligence and how we put guardrails around that and leverage some of the most modern techniques to enable it to work, I think for the most part, we’ve made the right bets,” he said. “This is hard, and you have to get all of them right.”
Read the full article here
