I was recently told that cars “are so 2016” by my esteemed, future-obsessed colleague, nicolas debock, but on yesterday’s announcement of Intel’s $15Bn acquisition of Mobileye I figured I’d give a throwback to 2016 and open up an internal memo I sent around Balderton on the automotive software market last year. [For a far more digestible version, it was also the base for an article in VentureBeat last October]. Some things will have changed since then but for anyone interested in the market and with some time to spare it should be a good round-up of what led to to Intel getting involved.
Albeit in memo form and looking at the market generally and from an early stage investor’s perspective… Sorry.
___________________________________________________________________
Overview: Investing in automotive software start-ups
1. Market — huge shift in the automotive market, as suppliers are gaining influence and OEMs are acquiring to stay relevant in the face of shared ownership and automation. Few markets have the amount of potential acquirers and recent M&A activity, as well as a trend towards acquisitions in software / tech specifically.
2. Tech step-change — deep-learning elements needed for self-driving are a great equaliser in the market today and existing suppliers and manufacturers alike lack the machine-learning and software development capabilities they need to fully pull this off. We have seen a number of start-ups enter the market in the past few years and several groups are in the race to take advantage of this shift from the supplier side:
… Tier 1 ‘mega-suppliers’ like Delphi Automotive, Valeo, Bosch, Denso, Autoliv and Continental are likely to build a presence, and ADAS-pureplay vendor MobilEye is positioned to do well.
… Chipmakers like NVIDIA are already building partial software solutions, mainly around vision and perception, to own a chunk of the market and continue to sell their chips — Qualcomm, Intel and TI may very well be up to something similar.
… New players supplying OEMs are popping up and getting funded, including Cruise Automation (acquired by GM), Comma.ai, Drive.ai in the US and Five.AI, Oxbotica, Adasworks here in Europe.
3. Time to predictable / recurring revenues — for early stage companies in automotive software, revenue will be POC/project-based for a long time, before annual license fees and royalties per unit sold kick in. Building meaningful revenue from autonomous vehicles specifically will take time, given the market is still immature.
4. Working with large and tricky customers — OEMs and tier 1 suppliers are difficult to work with both due to their size / length of their sales cycles and because they pressure they pass on liability / warranty claims to their suppliers. To minimise risk they either prefer to build and test in-house or use a restricted base of preferred suppliers. Building a company that can withstand the liability issues will require a large amount of capital.
Market size
Most sources estimate total of revenues in the Advanced Driver Assistance System (ADAS) market to be between $5 billion and $8 billion in 2015. Growth rates are consistently estimated above 10% and often between 15–30% annually, making it either one of or the fastest growing segment in automotive. ADAS has become incredibly strategic to OEMs, which has driven significant M&A activity in the area.
Technological change — As we move from drive assist (Level 1 and Level 2) to semi-autonomous and fully autonomous driving (Level 3 and Level 4), software makes up a bigger proportion of the car’s cost/value chain. 10 years ago the cost of electronics and software content contributed to 20% of total costs of a car — today that has risen to 35%, and more importantly has been seen to contribute more than 90% of innovations and new features.
The hardware is pretty much there — sensors and vehicle-to-vehicle / vehicle-to-infrastructure technologies that have been developed to date do not require huge additional technological leaps for autonomous driving and existing computing platforms should be able to deliver the necessary processing power. Costs still need to come down to make mass-market models.
In order to deliver semi-autonomous and fully autonomous driving, however, the software used to determine a car’s position and trajectory on the road must deliver a new level of capabilities compared to the current ADAS software that is on the road today.
Sensor fusion becomes more important as cars will need 360 view for Level 3 and 4, and will likely need multiple cameras, Lidar and Radar combinations. Today, ADAS systems like MobilEye tend to use only one forward-facing camera with a 50-degree field of view.
Overall, reports indicate that these shifts will lead to the ‘creation of very large markets for components/software services that barely have sales in the automotive space today’ and will mean that OEMs will need to work with suppliers outside of traditional automotive.
Pressure on car manufacturers — several trends are adding pressure on the OEMs’ influence in the value chain in the longer term:
1. …shifts towards shared mobility (car hailing, car sharing, ride sharing etc.) mean a lower rate of growth in unit sales in the longer term (McK estimate it will drop from 3.5% p.a. to 2% in the next 10–15 years).
2. …as technology and software become a larger part of a vehicle’s value, the supply base will have a larger and larger influence on the product.
3. …new entrants like Tesla, Apple, Google and Baidu , which are used to dealing with short product cycles and focus on software are challenging traditional players.
OEMs are aware of the challenges and taking action to protect their relevance through M&A, investments in new models, and more focus on decoupling software and hardware for faster development / release cycles.
Exit markets
Overall — Few markets have the amount of potential acquirers that we find in automotive — large companies, strong balance sheets after 5 years of strong profit in the industry, trends towards consolidation and major strategic shifts driven by technology and new business models.
M&A activity — was incredibly high in 2015. The total value of automotive-supplier deals in 2015 and 2016 was $74.4 billion (data compiled by Bloomberg) — far exceeding the $17.7 billion annual average in the previous 10 years. The number of transactions valued at $500 million or more rose to 18 in 2015 (3x the average for the previous 10 years), and there have been 11 in 2016 YTD. 2016 has seen a slow-down from its peak in 2015.
Acquisitions have been driven by multiple factors — global auto manufacturing expansion, new vehicle technologies, regulatory requirements, an abundant capital pipeline, and lucrative opportunities for profitability and growth.
Software- / autonomous-related acquisitions
… Continental acquisition of Elektrobit’s automotive-software division in 2015 ($665M)
… ZF Friedrichshafen acquisition of TRW, which builds safety, steering, braking and autonomous driving technology,, in 2015 ($12Bn) . “We’re following a building block approach to automated driving functions showcasing what is achievable today using proven technology,”.
… Uber acquisition of Otto in the same year as it was founded, allegedly to get access to the ex-employees from Google, Apple, Tesla working at the company ($680M)
… Harman International Industries acquisition of 2 software companies in 2015 ($780M for Symphony Teleca, a software services company, and $170 million for Red Bend Software, connected devices and over-the-air updates)
… GM acquisition of Cruise Automation in 2016 ($1Bn)
…Toyota puts together a deal to hire the full 16 people team from Jaybridge Robotics and invests $1Bn into the Toyota Research Institute
… Intel acquires two companies, Yogitech and Itseez, for their ADAS, robotics and autonomous machines division in 2016
Competition in ADAS
A. Direct Competition — supplying the automotive market
1. Tier 1 suppliers: MobilEye, Delphi Automotive, NVIDIA, Valeo
2. Competing tech companies: Google, Drive.ai, Cruise Automation (acquired), Oxbotica, Adasworks, Comma.ai, Five.ai.
B. Indirect competitors — building vehicles
1. Technology companies: Apple, Uber, Zoox
2. OEMs: Tesla, GM, Daimler, Toyota, BMW, VW etc.
A. Direct competition
… Driver assistance (ADAS) vendors — ‘mega-suppliers’ like Bosch, Denso, Autoliv and Continental providing hardware and software for multiple parts of the car, pure-play ADAS vendor MobilEye ($9.9Bn market cap)
… Chipmakers building partial solutions (e.g. vision algorithms etc.) — NVIDIA, Qualcomm, Intel, TI
… New players supplying OEMs — Cruise Automation (acquired by GM), Comma.ai ($3M investment from A16Z), AdasWorks, Drive.ai ($12M from undisclosed investors, Stanford research spin-off), Five.ai.
B. In-direct competition
… Large tech companies building in-house solutions — Apple, Google, Baidu, Uber, Tesla (previously used MobilEye)
… New players building vehicles — Otto (acquired by Uber), NuTonomy ($24M investment from Samsung Ventures, Highland Capital, Fontinalis Partners; spin-off from MIT), Zoox ($220M investment from Aid Partners Capital Holdings, Blackbird Ventures, Lux Capital, DFJ; team from Stanford that worked on Google’s self driving car project),
… Traditional car manufacturers building in-house — GM, Toyota, etc.
… Traditional car manufacturers currently partnering —e.g. Ford
Nearly all large car manufacturers have committed to delivering self-driving cars by 2020, but it is unclear that they have the expertise internally. Many seem to worry about working with the leader in the space, MobilEye, specifically due to a) lock-in (combined cameras / chips / software) and b) lack of products above level 1–2 automation (a single control function or multiple combined control functions).
Company details
A. Direct Competition — supplying the automotive market
1. Tier 1 suppliers
MobilEye (NL/ISR) — current leader in camera-based ADAS, building optical vision systems with motion detection algorithms. Current products are only level 1 and level 2 (single control function or combined) and require the use of their own chip (EyeQ) — plans to get to semi-autonomous driving (level 3) by 2018. Tesla recently stopped working with them for their Autopilot feature.
· Founded: 1999
· Market Cap: $9.9Bn
· Revenue: $241M (+68% YoY)
· Employees: 543
Delphi Automotive (UK) — one of the world’s largest auto parts manufacturers. Also building autonomous vehicle software to run on custom computers in the car. Completed a cross-country trip using autonomous technology in 2015 and testing autonomous vehicles in Singapore and expect to have fully autonomous vehicle (level 4) ready for 2018/19.
· Founded: 1994
· Market Cap: $17Bn
· Revenue: $17Bn (2014), declining in 2015
· Employees: 160K
NVIDIA (UK) –released the DRIVE PX 2 autonomous car platform (SOC), combining deep learning, sensor fusion, and surround vision, along with the DriveWorks SDK. The first-generation was delivered to 50 automakers, tier 1 suppliers, developers and research institutions last year. NVIDIA’s platform is a suite of software tools, libraries and modules that accelerates development and testing of autonomous vehicles, not a full software stack for self-driving.
· Founded: 1993
· Market Cap: $32Bn
· Revenue: $17Bn
· Employees: 9.1K
2. Select competing start-ups
Drive.ai (US) — developing a sophisticated automated driving system that it will eventually offer to carmakers. Uses deep learning for more elements of automated driving, including image recognition and elements of motion planning and control. Permit to test autonomous vehicles on the roads of California.
· Founded: 2015
· Funding: $12M
· Founding team: former robotics researchers at Stanford’s Artificial Intelligence Lab
· Employees: 11
Cruise Automation (US) — acquired by GM for $1BN. Angle was to convert existing cars into self-driving cars with three component of the system: software, sensor units on top of the car and a computer in the trunk.
· Founded: 2013
· Funding: $19M
· Founding team: former MIT student and founder of Twitch
· Employees: 40
Comma.ai (US) — hacker that wants to deliver aftermarket autonomous drive technology by the end of 2016 at low cost and using standard components Currently focusing on collecting as much driver data as possible. Investment from A16Z.
· Founded: 2016
· Funding: $3M
· Founding team: George Hotz, first person to hack an iPhone, built a self-driving car in his garage a month’s time.
· Employees: 4+
B. Indirect competitors — building vehicles for own use / sales
Uber — set up the Uber Advanced Technologies Center and hired away 40 researchers and faculty from Carnegie Mellon’s National Robotics Engineering Center. Acquired Otto, a driverless truck start-up, for $680M earlier this year (founded in the same year, with 91 employees including former employees from Google, Apple, Tesla). Partnering with Volvo to launch self-driving cars that can be hailed in Pittsburgh as of September 2016.
Google — Google’s autonomous vehicle project uses a spinning range-finding unit (Lidar) on top of the car with 64 lasers and receivers. The device creates a detailed map of the car’s surroundings as it moves, also adding information from other sensors and compares the map with existing maps (dense 3D point clouds), alerting the system to any differences. Because it relies on existing 3D maps that need to be continuously updated, it is limited to mapped routes and the cost of LIDAR’s remains high — Google’s kit from company Velodyne rumoured to costs $75K.
Apple — have never confirmed an autonomous vehicle project, but have hired people from companies like Tesla, Ford, Mercedes and Toyota, and Tim Cook has mentioned that they are moving into cars generally.
Tesla — 90,000 Teslas already have Autopilot, which allows the cars to accelerate, maintain lane position, change lanes and park, developed using Mobileye chips and software. With a recent fatality and disagreements between the companies on how the software was deployed, they ended their partnership a few months back, the ex-head of supply-chain at Tesla confirmed that they had been phasing out MobilEye for years and that they are disliked by most of their customers. Tesla expects to have a fully autonomous car by 2019.
· Founded: 2003
· Market Cap: $33.2Bn
· Revenue: $4Bn
· Employees: 13K
Zoox (US)– “stealth” company building a full vehicle, intended to compete against Uber. Does not appear as though they will be selling software to existing car manufacturers. Permit to test autonomous vehicles on the roads of California.
· Founded: 2014
· Funding: $240M
· Founding team: Tim Kentley-Klay (designer) and Jesse Levinson, a Stanford engineer who worked on self-driving cars with the co-creator of Google’s self-driving car
· Employees: 140
