Experimental feature

Listen to this article

00:00
00:00
Experimental feature

General Motors celebrated being the world’s largest carmaker for the 76th straight year in 2007. It was sitting on $25bn in cash. Eighteen months later, it was bankrupt.

The automotive industry is among the most capital-intensive in the world: If the economy sours, assets turn into liabilities overnight as factories churning out thousands of cars begin to haemorrhage cash. So when toxic mortgage securities blew up in 2008, causing a recession, banks performed terribly — but carmakers fared even worse.

That is what makes auto consultants at Bain so worried. They fear that carmakers are about to be hit with a one-two punch: first, they project a US recession in the next 12 to 18 months. Then, increasing numbers of baby boomers will retire, causing a structural decline so big that, they warn, US car sales could shrink from more than 17m last year to just 11.5m by 2025 — the same level seen in 2008-09, which caused GM and Chrysler to go bankrupt and Ford to suffer a $14.6bn loss.

“The collapse in auto sales in the coming years could be as severe as it was during the great recession,” says Bain partner Mark Gottfredson. “Only this time, the sharply lower demand will be permanent.”

column chart comparing the value of selected car manufacturers with the new players in the autonomous car market

But there is hope. If carmakers play their cards right, they could be saved by what GM has called “the biggest business opportunity since the internet”. The potential saviour is the rise of shared, driverless “robotaxis”, which Bain expects to become mainstream in some large cities in six to eight years.

This new market, virtually non-existent today, promises to be huge. Analysts at UBS estimate that its revenue in 2030 will be between $1.3tn and $2.8tn, a forecast based on robotaxis accounting for 12 per cent of new car sales. By 2050, when they are likely to be far more common, chipmaker Intel projects a “passenger economy” worth $7tn. Global vehicle sales stand at $2tn today.

Car brands typically earn $2,000 from a vehicle sale. That is just $0.01 per kilometre over the lifetime of a vehicle, whereas for robotaxis “the potential is 20 to 25 cents per kilometre”, says Andreas Tschiesner, automotive lead for Europe at McKinsey. “So there is huge potential to capture more.”

To realise this potential the industry will need to update its entire business model. The challenge for carmakers is to gain the expertise in self-driving algorithms, in-car entertainment, streaming services and fleet management for ride-hailing that will be central to this new era.

A Cadillac fitted with GM's Super Cruise technology. When the light is green on the steering wheel, Super Cruise is active. Sensors prompt the driver to turn their attention back to the road if they have been looking away for too long © GM

Luckily, there has been an explosion of small companies developing the skills and technologies that carmakers can make use of. According to McKinsey, $211bn has been invested into mobility start-ups since 2010. Only 7 per cent came from the carmakers. But the majority was financed by venture capital and private equity funds, creating a swarm of small players that have an incentive to sell their breakthrough technology up the traditional value chain.

The risk is that the carmakers fail to integrate this new technology. Citi analyst Itay Michaeli warns of “an industry race the likes of which we haven’t seen before”. If the carmakers fail, they could find themselves relegated to the status of a supplier. Or worse. “Many may not survive,” says Mr Gottfredson.

Waymo, the Alphabet self-driving unit that began as a Google project, is widely seen as the leader in this new landscape.

On two key indicators — miles driven autonomously and “disengagements”, or the average number of miles driven without human intervention — it has built a commanding lead since its founding in 2009. And with at least 600 of its vehicles driving more than 25,000 miles a day, it is perfecting its algorithms in a way that could blindside the competition. Last year UBS projected that Waymo “will dominate” the operating systems for autonomous vehicles, taking “60 per cent of the total projected revenue pool in 2030”.

Banks are already giving it sky-high valuations. In 2017, before Waymo had even earned a dollar of revenue, Morgan Stanley valued it at $70bn — roughly the same as Volkswagen, the world’s largest carmaker by sales. Last year the bank realised it had not taken into account the potential for Waymo to license its technology and enter logistics, where it could help Walmart deliver goods to better compete with Amazon, for example. It revised its valuation to $175bn.

Last month, Jefferies went further. On the assumption that Waymo can take a 2 per cent share of all miles driven worldwide in about 10 years, it valued the group at $250bn. That is more than Ford, GM, Fiat-Chrysler, Honda and electric carmaker Tesla combined.

Mighty AI uses crowdsourcing to help improve the identification skills of autonomous sensors and processors by tagging images from the real world © Mighty AI

“I firmly believe that within five years the majority of carmakers will come to Google and say, ‘we need your help’,” says Brent Thill at Jefferies.

The threat of Waymo is not that it will build better cars. It has no need to. Instead it is ordering vehicles from Chrysler and Jaguar — effectively turning them into suppliers — and then fitting them out with self-driving software and hardware built in-house.

But its potential goes beyond superior self-driving capabilities. Once robotaxis are mainstream, Alphabet can collect data from Google Maps and Search, entertain with YouTube and the Play Store, offer advice through Google Home smart speakers and use its software knowhow to manage fleets. Aside from the vehicle itself, Waymo is a vertically-integrated “closed system”, says UBS.

“This will influence advertising, the media and the entertainment business,” Mr Thill says. “It’s not just the autonomous technology, it’s all the components that Google brings to the car. This is why it is putting so much investment into the living room, because it wants the car to feel like your living room.”

A Waymo Jaguar i-Pace fitted with self-driving technology

Carmakers are scrambling to respond. They have partnered up like never before and made big investments to acquire new expertise. Volkswagen has linked up with Ford, while arch-rivals BMW and Mercedes have pooled their mobility efforts. In 2016 GM paid $500m for a stake in Lyft, the ride-hailing group, and it spent more than $1bn to buy Cruise, a self-driving company.

Bain’s Mr Gottfredson says the Cruise acquisition had looked expensive for a start-up with fewer than 50 employees. But with Japan’s SoftBank and Honda since buying stakes, its valuation has ballooned to $14.6bn. “Today the value of Cruise is underpinning the entire value of General Motors,” he says.

These deals, however, are merely the tip of the iceberg. Beneath the car brands, an entire ecosystem of niche companies has spurred into existence. Known as the “data value chain”, these groups specialise in the software, sensors, data processing and navigation needed to make autonomous cars a reality. None has the willpower, resources or vision to take on Waymo. Instead, they are forming clusters, exercising “swarm intelligence” to independently work towards the same collective goal of creating a safe, driverless experience.

New players include Israeli companies such as Iguazio, which specialises in real time data processing for ride-hailing apps, and Foresight Automotive, whose four-camera system, it claims, can detect obstacles in all weather conditions, regardless of whether the object has been seen before. “If an alien walks out of a UFO, we’ll detect it,” says Foresight vice-president Doron Cohadier.

The implications of this ecosystem are profound. It suggests the carmakers can catch the likes of Waymo up without being the best-in-class in the new technologies. They merely need to be competent enough to know who is best — and then partner with them.

A driverless shuttlebus at Mcity, a testing ground built by the University of Michigan with funding from car companies including Ford, GM and Honda, for testing driverless vehicles © University of Michigan

“This is why they have a very good chance to win this fight,” says Adi Pinhas, chief executive of Brodmann17, an Israeli start-up that uses machine learning to cut down the computing power needed to digest data. “We just do data processing, others just do mapping, others do sensing. When you have these large, specialised teams working on these building blocks, and the carmakers are at the top orchestrating, they get faster progress.”

Bob Lutz, former vice-chairman of GM, adds: “I believe the huge virtual network of thousands of little companies will ultimately deliver the goods, if people know how to take advantage of it. Collectively, they will overpower any monolith.”

Mighty AI, a Seattle-based start-up with 85 employees, helps autonomous vehicle cameras and sensors to “make sense of the raw data” by detecting objects and labelling them. What is unique is that it takes real-world footage and then crowdsources the mundane task of tagging images to more than 500,000 people who use their app. “To label one image might be 50 tasks,” says Daryn Nakhuda, chief executive. “That isn’t something you want your data scientists, computer visualisation teams or manufacturers thinking about.”

Via, a US ride-pooling app founded in 2012, has collected data from 50m rides worldwide. This has enabled it to match cars with passengers quickly, and make sure they are fuelled or charged at optimal times. It has partnered with Daimler in Europe.

“We are uniquely focused on this one thing,” says Daniel Ramot, Via’s founder. “It’s a problem we have been solving the last six years. It’s quite complex. Sometimes the autonomous vehicle guys ignore all this and just think about the algorithms.”

When Google’s self-driving car project began a decade ago, this data value chain did not exist, but it is emerging fast. McKinsey’s database of mobility start-ups includes 1,180 companies. Israel alone is home to more than 400, including Mobileye, the vision group Intel paid $15.3bn for in 2017. Mobileye is developing an open platform for autonomous driving with help from BMW, Fiat-Chrysler and suppliers Delphi and Magna.

The various joint ventures, partnerships and takeovers are “combinations which you could not have imagined just a few years ago”, says Axel Schmidt, head of global automotive at Accenture. “Everyone has realised there is no cluster on its own to offer this broad mobility game.”

The Israeli startup Mobileye, now part of Intel, is working on an autonomous ride-hailing system in partnership with Volkswagen and looking at selling its technology to truck companies

Parts of this emerging network have already helped GM gain wide acclaim in the industry for Super Cruise, its hands-free driving technology. A key enabler for this service came from Ushr, a company with 60 employees which provided GM with high-definition maps covering every major highway in Canada and the US.

Ushr mapped highways to an accuracy of 15cm using laser-based lidar (light detection and ranging) technology. While other self-driving technologies have terabytes of data in the cloud that gets transferred to the car when needed, Ushr reduced 130,000 miles of high-definition maps to a 300-megabyte file that stays in the car, says Chris Thibodeau, Ushr general manager.

“GM has approached autonomous driving from a systems integration perspective,” Mr Thibodeau says. “It’s not about the technology of one company, it’s about taking data from the camera, maps, the sensor, and making it work well together.”

Var chart showing the estimated sources of revenue in autonomous car manufacture in 2030

This ability to select the best partners and integrate different technologies is a major advantage for the carmakers, says Thomas Müller, head of development for autonomous driving at Audi.

“There is not just one company that can do everything. You can follow that approach, but it will be expensive and risky,” Mr Müller says. “We learnt how to work in a network, to manage that network of different partners, and to integrate that into the car, a product or a piece of software.”

Dan Glotter, chief executive of OptimalPlus, a big data analytics company, says the traditional pyramid structure of the value chain is giving way to a hub and spoke model in which carmakers sit in the centre and interact with all levels, including the makers of semiconductors and radar systems.

“Suddenly the carmakers need to speak with their tier-two, tier-three suppliers, not just the Boschs and Continentals,” he says. “There is now a different level of complexity.”

Mr Glotter says this gives carmakers an advantage, because playing the role of integrator is what they have been doing for decades. “BMW has an order of magnitude more experience dealing with the supply chain [than software companies],” he says.

Ronny Cohen, chief executive of Israeli start-up Vayavision, which builds 3D models of a car’s surroundings by fusing raw data from the camera, lidar and radar systems, says working with carmakers is the goal for suppliers. “My business model is the licensing fee,” he says. “They manufacture the cars of the world today, so they have the volumes.”

This approach can be dismissed as “outsourcing innovation”, but the benefit is that the enormous risks and costs for new technologies are shared, increasing the chance of success.

“Supply chains contain exponentially more expertise and resources than any single company,” says Alex Saric, chief marketing officer of procurement platform Ivalua. “So the most successful organisations will be experts at unlocking the innovation in their supply chains.”

Copyright The Financial Times Limited 2019. All rights reserved.

Follow the topics in this article