General Motors made headlines in March when it paid over $1 billion for Cruise Automation. A few weeks later leading venture capital firm Andreessen Horowitz entered the space, announcing investments in two early-stage autonomous startups,Comma.ai and Dispatch.
Most recently, secretive AV startup Zoox raised a massive $250 million funding round, making it Silicon Valley’s newest unicorn. These and other recent deals point to a growing investment frenzy as AVs get closer to mainstream commercialization.
The AV investment landscape is complex. It includes both hardware and software players and features competitors ranging from early-stage startups to large publicly traded corporations. This article will provide a primer for those interested in the rapidly evolving AV space.
The first and most obvious layer of the AV ecosystem is the vehicle itself. The capital investment and manufacturing expertise required to produce vehicles at scale largely preclude early-stage entrants from being active here. Even large, deep-pocketed technology companies investing heavily in an autonomous future — e.g. Google and Uber — seem unwilling and unlikely to become car manufacturers themselves.
The most probable outcome therefore seems to be that traditional car manufacturers will continue to mass-produce vehicles in the autonomous age. It is unclear whether this manufacturing role will continue to be as profitable for these companies as it has in the past. As value creation in transportation shifts toward high-tech components and software, manufacturers of the cars themselves may become an increasingly commoditized, low-margin business.
Virtually every traditional car manufacturer has by now begun to invest in autonomous vehicle capabilities. Those with particularly interesting autonomous programs include GM, Volvo and Mercedes-Benz. (While Tesla manufactures cars, it is more appropriately considered a technology company.)
Lidar is one of several types of specialized sensors that allow AVs to interpret their environment. Lidar sensors give the vehicle a precise three-dimensional awareness of its surroundings by projecting lasers in all directions and measuring the time they take to rebound, a process analogous to radar (the word Lidar is a portmanteau of “light” and “radar”).
Given how critical these components are for overall AV functionality, the market for Lidar sensors will be enormous. A handful of startups have recently emerged that specialize in their production.
Two key dimensions of these sensors are their size and their cost; the company that can harness Moore’s Law to drive both of these down the fastest will have a huge advantage. The Lidar sensors that Google used for its initial AV prototypes reportedly cost $80,000, an impracticable price point for the mass market.
Another California-based Lidar startup that has attracted positive attention recently is Quanergy. Quanergy has announced that its sensors will cost only $250 and have no moving parts. The company has established relationships with a handful of large OEMs but has yet to bring a product to market.
Like Lidar sensors, cameras help AVs understand their environment and maneuver accordingly. Though less precise than Lidar, cameras offer the significant advantage of being able to detect color—important when, for instance, identifying traffic lights and signs.
The dominant player in AV camera production is a publicly traded Israeli company namedMobileye. Mobileye has high-profile supplier contracts with a number of auto manufacturers including Tesla.
Perhaps the most important piece of AV hardware is the computer chip that serves as the vehicle’s “brains.” These chips take inputs from the vehicle’s various sensors and, based on complex software algorithms (discussed further below), enable the vehicle to operate autonomously. As with microprocessors in personal computers, these components sit at the very center of the overall system’s functionality.
Given the enormous computing power demanded, AVs will require state-of-the-art microprocessors. The established chipmakers that have long dominated the microprocessor market — Nvidia, Qualcomm and Intel — seem poised to leverage their existing expertise to succeed with AV chips. All three companies have signaled that autonomous technology will be a strategic priority moving forward.
Of the three, Nvidia is arguably taking this opportunity seriously and investing most heavily in it. Investors have taken notice, with the company’s stock trading near an all-time high.
While the hardware described above is essential, AVs are able to act intelligently, or autonomously, because of their software. There are several different key types of AV software to be aware of.
It is worth noting that the divide between hardware and software companies, while helpful as a framework, is not entirely clean. Some hardware companies — e.g. Mobileye — also provide software to analyze their sensor data. Likewise, some companies classified below as software players also offer hardware as part of an end-to-end autonomous solution.
Mapping and localization
The first category of software critical to AVs is mapping and localization.
In order to effectively navigate, an AV must have a detailed and up-to-date map of its surroundings and must know where on that map it is located. Creating and continually updating such a map database is a massively challenging exercise.
The two biggest players specializing in digital worldwide map database creation are HERE and TomTom. Each of these companies has attracted significant investment attention — no surprise, given that maps will be a key strategic asset for the AV industry.
A coalition of German automakers including Audi, BMW and Daimler recently acquired HERE for around $3 billion (outbidding Uber, among others).
TomTom, a publicly traded company based in Amsterdam, has faced acquisition rumors for years by suitors including Apple; to date the company remains independent. Apple, Uber and Bosch all have partnerships to use TomTom’s data.
Meanwhile, other AV players — notably Google and Uber — are seeking to build mapping capabilities themselves, while a handful of smaller startups are also tackling this challenge.
As vehicles become increasingly connected to the Internet, other vehicles and surrounding infrastructure, cybersecurity will become an increasingly prominent concern. In an important warning of the potential dangers of connected vehicles, white-hat hackers last year remotely took control of a Jeep Cherokee and cut its transmission.
Entrepreneurs and investors are becoming active in AV cybersecurity. As examples, Tel Aviv-based Argus Cyber Security raised $26 million in Series B funding last year, auto electronics maker Harman paid $72.5 million for cybersecurity startup TowerSec in March, and newcomerKaramba Security raised $2.5 million in seed funding in April. More competitors will no doubt emerge soon.
Fleet operations and management
As the autonomous age dawns, many predict that private car ownership will become obsolete, replaced by shared AV fleets that individuals summon only when needed. The task of managing these fleets and optimizing their routes will be an immense challenge requiring complex software solutions.
Startups already beginning to tackle this challenge include RideCell, which in early April raised $11.7 million from BMW and Khosla Ventures. Given its strategic positioning and its commitment to autonomous technology, it seems safe to assume that Uber will invest and compete vigorously here.
AV artificial intelligence / machine learning
The central technological breakthrough at the core of the entire AV concept is the vehicle’s ability to conduct advanced and adaptive decision-making itself based on all the data at its disposal. Artificial intelligence software enabling vehicles to “think” in this way is the most important and technically demanding AV technology category of all. A handful of companies are seeking to build such solutions.
Some of these companies focus solely on software; to go to market, they will look to partner with, or be acquired by, hardware manufacturers. One prominent example is nuTonomy, which recently announced a partnership with the Singapore government to deploy driverless taxis there by 2018. nuTonomy, an MIT spinout, raised $16 million in Series A funding in May.
Other companies are building machine-learning software integrated with hardware in order to offer a comprehensive autonomous system. Included in this group are auto manufacturers such as Tesla but also many promising startups. Cruise Automation, recently acquired by GM, is one well-known example.
Another noteworthy startup is George Hotz’s Comma.ai. Comma.ai is building aftermarket “kits” consisting of sensors, computers and software that allow customers to convert existing cars into AVs. The company aims to bring these kits to market for under $1,000 by the end of 2016.
The landscape of AV companies in these early days of the technology is fluid and fast-changing. Established auto manufacturers, large technology companies and scrappy start-ups are all fiercely competing to win in the AV ecosystem.
A wave of M&A activity, partnerships and consolidation seems likely as AVs move toward commercial availability. One thing is certain: There will be massive opportunity for profit as the autonomous vehicle market takes off in the coming years.
In the words of angel investor Tikhon Bernstam, one of Cruise’s earliest backers: “You’d be hard-pressed to lose money investing in this space right now because there is going to be tens or even hundreds of billions in M&A and IPOs going forward. Cruise at $1 billion may look very cheap one day.”