Combustion is The New Cartographer's guide to the
present and future of the auto industry.
We provide insight via our email newsletter to executives, investors, and decision-makers on where technology, policy, and the market are headed next.
When GM and Ford made their respective billion dollar bets on Cruise and Argo, they weren’t buying self-driving car technology. Neither of the startups had developed any proprietary technology at the time that couldn't be matched by GM or Ford’s internal efforts. Rather what GM and Ford were buying was talent. They were bets on who the investment and capital structure could hire and retain, the reboot necessary in order to make a long bet on a self-driving future.
But this central challenge– retaining world-class software talent– is going to create an existential threat to OEMs far before either’ ability to deliver the goods on a fully self-driving car is called into question. This is because the fundamental change that enabled a credible path to autonomous vehicles was an underlying soup of possibility that had little to do with cars: the combination of computing power, algorithms, and data that first became available to Google, and has now started to become accessible to companies of all shapes and sizes to apply machine learning to nearly any field.
Autonomous vehicles weren't enabled by a radical new car-specific sensor hardware. It was the opposite: the hardware supply chain is moving to innovate to match the opportunity provided by software and software infrastructure. And this software ability– fundamentally, the ability to interpret historic data to make predictions– is going to have implications in every industry.
For auto manufacturers, this means that machine learning will begin to play a role in:
1) Designing vehicles. Can we reach higher levels of safety and efficiency by using machine learning to run many scenarios and pick 'winners' for further exploration? Can the design cycle and testing of vehicles be shortened? This is an area that experts will be particularly hesitant to believe there are increased efficiencies. But Google found it a useful tool in their own backyard, when they found that machine learning could reduce data center cooling by 40% – a core and important part of their business for which they had previously employed the very best experts in the world. The same will be true about every part of safety and efficiency design of the automobile.
2) Optimizing the supply chain. Machine learning can play an important role in prioritizing requests and deliveries based on historic demand and indications of future interest. These changes will result in vehicles being built and delivered to customers more quickly than ever, with less waste. Expect machine learning to also optimize quality assurance, allowing more efficiency checking while improving output.
3) Interacting with customers. Machine learning should play a role in creating a customized interaction with each customer at each point they touch OEM's brand, whether browsing models online, being directed to the right dealers or interacting with customer service after a purchase is complete. The ability to customize at scale should increase short term revenue (by improving the effectiveness of marketing), long term revenue (by improving customer loyalty), and cut costs.
This opportunity for OEM poses a real challenge, for they will have to compete in the same machine learning labor market to work on core business processes, a far less attractive problem than the autonomous vehicle. This will mean evolving brand, work practices, and increasing flexibility. And if auto OEMs can hire this talent– it will be expensive. As one data point, Chinese giant Baidu spent $2.7 billion over the last two years to employ 1,300 researchers focused on machine learning.
Auto OEMs will have to nimbly examine their core businesses lines and processes to understand where they choose to differentiate, and where they choose to outsource to the evolving machine learning technologies from companies like Microsoft, Amazon, Google, and IBM. But no part of the business should remain unexamined or auto OEMs will face fiercer competition far ahead of when autonomous vehicles arrive to market. Automation will hit the the factory and balance sheet far before it hits the road.
Sam Gerstenzang was mostly a principal at Alphabet’s Sidewalk Labs, where he cofounded the transportation data company Flow. He previously ran product, data, and design for Imgur, and was a member of the investment team at Andreessen Horowitz.