Startups, corporates, and technologies that are shaping the future of autonomous driving
From 2014 to 2019, funding for the auto and transport industry was US$70Bn. Meanwhile, autonomous driving industry got US$19.3Bn in funding. Although the funding is at an all-time high, seed funding rounds are at an all-time low and this trend will continue next year. As a result, we will see fewer new startups entering this space.
Partnerships and M&A are on the rise and this trend will continue well into next year. And, rather than competing, incumbents are collaborating with startups to offset the cost and share resources. With fewer market gaps and a huge technical barrier for entry, the autonomous driving industry is quickly turning into a race towards autonomy.
So, I decided to tear this industry down and analyze startups (and deliberately look beyond Tesla and Comma.ai), corporates, and technologies that are shaping autonomous driving and get answers to these questions.
- What is autonomous driving? — Five levels of autonomous driving
- What has happened? — To answer this question, we looked at the money, tech and corporate interest that has been driving the industry.
- What’s happening now? — We analyzed 270+ companies that are working on components for AVs, full-stack self-driving solutions and fully autonomous vehicles to paint the landscape.
- What’s next? — Nine trends that are shaping the future of autonomous driving
What is autonomous driving?
There is a lot of literature that can explain autonomous driving. To quickly summarise, AV or Autonomous Vehicles sense the environment and self navigate safely with little or no human input. These vehicles use sensors, lidar, sonar, GPS, and cameras to detect the environment and navigate using built-in advanced control systems.
According to the Society of Automobile Engineers (SAE), autonomous driving can be categorized into five levels.
Right now, the big auto manufacturers are in levels 1 and 2, while startups and tech companies are in levels 2 and 3.
What has happened?
We have been trying to build autonomous vehicles since the 1920s (Phantom Motor Car). But in the last five years, money, tech, and interests aligned to make this a reality. So, let us look at this one by one.
Money
Funding is at an all-time high. In 2018, the Autonomous driving market saw a US$7.23Bn in funding across 141 deals. This is an 86% increase from 2017, and this is a 4.5x time increase from 2011.
If you look at the deal share by stage, 64% of deals made in 2018 were early-stage funding rounds (seed and Series A), but seed rounds have decreased considerably this year. 17% of the deals made this year are seed rounds when compared to 46% in 2016. It looks like the number of seed rounds might go down further in 2020 as investors consolidate their investments from 2016 to 2018.
Also, we saw an increase in series A funding as it rose from 23% to 33% while late-stage funding has gone down by 2%. This year, late-stage funding is bouncing back and now holds 28% of the total deals.
The average funding deal hit a record high in 2019. After hitting a record low in 2016, the average deal size is on the rise for the last three years. In 2018, the average deal size rose by 81% from US$40.05M to US$64M, and now it is at US$122M.
Median deal size climbed 7% from US$10.66 in 2017 to about US$18.9M this year.
Here is the list of top-funded companies in autonomous driving
To summarize, funding is at an all-time high and will continue to grow at the same pace next year. Meanwhile, seed funding will continue to decline further.
Tech
To see the evolution of tech in this industry, we analyzed over 800 patents (using google patents) to figure out what has happened.
The number of patents filed patents peaked in 2017. 321 patents filed by 118 players/holders in the autonomous driving market.
591 patents were focused on steering control, decision making, situational awareness, and more.
We categorized them as autonomous cars. Another major category was Adas/Advanced driver-assistance system and 216 patents well in this category.
Toyota holds the maximum number of patents (127 patents) followed closely by Baidu, Intel, and Waymo.
Interest
We can split the interest into two parts — corporate interest and news coverage. For corporate interests, we analyzed over 600 earning calls in the last five years.
Corporates talked about their plans for autonomous driving to their shareholders 1388 times in the last five years, and the interests peaked in Q1 of 2017 with a slight decrease in Q1 of 2018 and 2019.
BMW loves autonomous driving, and its top executives talked about autonomous driving more than 100 times in the last five years. Here is a list of corporates and the number of times they spoke about autonomous driving
- Mobileye — 95 times
- NVIDIA — 80 times
- Visteon — 57 times
- TomTom and Diamlet — 52 times
- Intel — 42 times.
In the news, mentions about autonomous driving peaked in Q1 2018 due to uber’s Arizona accident, but the overall trend line is going up.
Now that we have caught up with what has happened in the autonomous driving industry. Let’s look at what is happening now.
What’s happening now?
To clearly understand what is happening now, we analyzed over 270 startups that are working on fully autonomous vehicles, full-stack self-driving solutions, and components (like sensors, mapping, simulation, computer vision) and more.
Autonomous Driving Systems
These are startups that are building a full-stack autonomous driving system like hardware and software (computer vision and sensor fusion software). They partner with automakers to deploy this tech, and in some cases, they retrofit existing vehicles as well.
AImotive, for example, offers software for cars and flights, and recently they partnered with NextChip to and announced aiWare, next-gen image hardware for self-driving vehicles. Drive.ai is another startup that partnered with Lyft to bring self-driving taxis.
China has unicorns in the autonomous driving industry. Pony.ai partnered with Guangzhou Automobile Group, China’s second-largest automaker, to bring autonomous driving fleets in Guangzhou. Momenta, partnered with the government of Suzhou to test and build a large-scale smart transportation system.
Complete Autonomous Vehicles
Companies like Faraday Future, Zoox, and Nuro are building AVs from scratch, and these vehicles are entirely different from the traditional cars on the road. These AVs don’t have steering wheels or dashboards. So, these cars are legally allowed to drive on public roads.
Nuro’s AVs are designed to carry cargo instead of people, and they target last-mile delivery. Faraday Future, on the other hand, develops fully autonomous electric vehicles with unique ownership models, in-vehicle content, and more. May Mobility is a Michigan based startup that develops AVs from scratch with a focus on system-level safety.
Mapping
Autonomous driving relies heavily on maps as they compare their surroundings to a digital map stored in their memory like humans. These maps are not google maps. These are HD maps with road-based information like lane sizes, crosswalks, and road signs. HD maps are built with data collected from sensors and software that turn them into a digital map.
Wayz.ai, a china based startup with US$80M in funding, offers HD maps, real-time location-based on cameras and sensors, safety testing, and cloud-based solutions for autonomous driving vehicles.
DeepMap is another company that develops map building software and licenses out to automakers and tech companies looking to teach vehicles how to drive. DeepMap got US$92M in Funding from Goldman Sachs, Bosch, NVIDIA, Accel, and A16z.
SLAMcore is another startup that develops an advanced algorithm to help AVs, drones, and other systems to simultaneously map the surroundings and position themselves within it.
Civil Maps is using AI to convert raw sensor data into meaningful maps.
When it comes to corporates, Apple acquired Coherent Navigation for its self-driving cars. Google is building its own HD mapping with Waymo, and Volvo has left TomTom for Google.
Salesforce acquired MapAnything, a startup that offers mapping, schedule planning, route optimization, real-time geo-location, territory management, and geo-analytics. Audi, Daimler, and BMW acquired HERE Maps. Baidu is building out its own self-driving platform called Apollo, and it is planning to monetize maps, and the largest company in China believes that HD maps will be larger than their current business.
LIDAR
LIDAR or Light detection and ranging use Infrared sensors to determine an object’s distance and sensors pulse at a rapid rate to measure the distance. Traditional LIDAR tech is expensive and uses a 360 spinning camera to capture the surroundings. So, startups in this space are trying to reduce the cost of lidar sensors while maintaining high accuracy.
Bajara, for example, built a spectrum-scan lidar for autonomous vehicles that uses prism-like optics and shifting wavelengths of light.
Innoviz uses solid-state lidar tech has partnered with BMW and Magna to get their scanners in the market. Velodyne has a relatively expensive lidar tech which houses 128 lasers. LeddarTech has a patented solid-state lidar and has partnered with Acal BFI to bring their lidar to the European market. Aeva claims to have a lidar team that has a range of 200 meters and shoots a continuous lightwave instead of individual pulses.
Oculii develops military grade 4D sensors and uses advanced sensor fusion techniques to manufacture smarter, high precision sensors and systems.
Camera and Computer Vision
The camera, along with ADAS, can spot road signs, traffic lights, and street markings, but it is not that great with depth perception and distance measurement. Elon Musk believes that you can do a lot more with a camera and that you don’t need an expensive lidar tech. So, cameras capture highly accurate images, and computer vision software detects objects, signals, lanes, and asses the appropriate traffic signs and rules.
Light, for example, builds cameras with 16 lenses to extract highly accurate 3D images for self-driving cars. SenseTime is another company that offers computer vision and AI for AVs, and NetraDyne is developing deep learning solutions and vision-based analytics.
DeepScale is developing advanced neural networks for computer vision. Meanwhile, Prophesee has built an event-based computer vision tech that mimics how the human brain processes images.
Radar
Autonomous driving needs radar to detect an oncoming object’s distance, range, and velocity. Radar doesn’t have spinning parts. So, it is more accurate than lidar, and costs are meager compared to lidar.
Echodyne, a Bill Gates funded company is combining radar with computer vision-like software to create 4D imaging for AVs.
Zendar develops high-definition radar for autonomous vehicles to navigate in bad weather. Lunewave is a startup that 3D prints antennas with greater range and accuracy. Metawave uses metamaterials like Echodyne for longer detection range.
V2X
Sensors can’t detect objects outside their line of sight. So, a new class of automotive sensors will allow vehicles to see what’s beyond the sight. As long as cars are connected to the same network, it can detect other vehicles, pedestrians, and traffic signals.
Autotalks is a semiconductor company developing VLSI solutions for V2X and V2I communication. Valerann has developed an IoT system to sense the traffic environment. Meanwhile, Peloton provides and manages tools for saving fuel, avoiding accidents, and improving operational insight through the use of connectivity, automation, and data analytics.
Data and Simulation
Autonomous Vehicles need to drive billions of kilometers to train the algorithms that guide the vehicle. This distance would take years. So, AV developers are amassing additional kilometers through simulation. With AI, startups are generating or augmenting existing datasets to train AVs, and this tech is useful in training AVs on dangerous and less frequent situations.
Startups like Cognata have developed a 3D simulation platform to provide customers with various testing scenarios. Parallel Domain, Righthook, and Metamoto are some of the exciting startups that help customers with simulation.
Now, what’s happening in with the autonomous driving industry, let’s look at what will happen next.
What’s Next?
To predict what will happen, we need to look at the data from a different perspective. So, I used data (like funding, momentum, customer adoption, media attention, and competition intensity) to evaluate each segment, business models, and their offering against market maturity and adoption to come up with three buckets.
Need to Focus
In this bucket, we’ll put in market segments that have seen a widespread industry adoption and market reach. The Incumbents and new entrants should have a clear strategy and initiatives for these trends.
ADAS, Radar & Camera — Until full autonomy, automakers will have to enhance their driver safety tech to work with humans to minimize errors. With advancements in sensor tech and computer vision, OEMs can bring in new features like adaptive cruise control, automatic braking, traffic and lane departure warnings to augment the driver’s capabilities and assist in case of distraction or fatigue.
For example, Toyota is trying to develop a vehicle that is “incapable of causing a crash.” They are trying to improve ADAS to monitor and assist humans behind the wheel.
Telematics and Driver Monitoring — Connected vehicle tech is useful in fleet operation and management. This technology provides substantial cost-saving opportunities for fleet managers, who can utilize telematics technology to track location data, fuel levels, and driver behavior. This allows them to optimize routes, minimize fuel costs, and increase fleet safety. Insurers love this tech and it is resonating with corporates across industries. According to McKinsey, monetizing telematics data could generate US$1.5T for automakers by 2030. GM, Volvo, and BMW are using connected car data to drive more informed insurance plans.
Should be working on
These following segments have significant market growth and considerable investment and partnering activity. Early adopters and corporates have embraced it, and these trends are on their way to gaining widespread industry and customer adoption.
HD Mapping — HD mapping will help AVs localize themselves with centimeter-level accuracy, but building such a complex map can be quite costly. It requires a fleet of vehicles equipped with sensors to capture roadway data along with the advanced infrastructure to process it.
Automakers are getting hands-on in this segment by acquiring mapping startups or partnering with competitors. Startups like DeepMap are trying to license their map building software as well.
LIDAR — Once considered a costly impractical tech, it has seen several improvements that are driving down costs and increasing reliability.
Some manufacturers charge over US$100,000 for a traditional unit (360 degrees spinning less reliable ones). Elon Musk thinks lidar is not necessary for AVs. Yet, it still powers Google’s Waymo. Plus, many startups are working to bring down the cost and increase the performance.
Price is not the only limitation here. Many lidar systems are not all-weather systems. It gets affected by heavy rains and low hanging clouds. So, lidar should be used as a complementary sensor to cameras and radars.
Fully autonomous driving vehicles — These are still years away from commercial adoption, but only a few (like Waymo and Drive.ai) have launched commercial services. Other services act as a backup for human drivers. Electrification has been on the top priority for automakers in the last few quarters. But investors have shown their confidence in the fully autonomous driving stack, and the initial testing ground is logistics (esp. Last-mile delivery).
Experiment with
These are trends in their early stages with a few function products/services and minimum market adoption.
Simulation — Driving simulation platforms help AV devs reduce the time and hassle associated with real road testing. With increasing AV devs relying on simulation to gather additional miles, NVIDIA took this opportunity and became the leading GPU chip provider for Tesla, Volvo, Volkswagen, and Daimler to run their simulations. Also, they have a cloud-based simulation platform DRIVE Constellation.
Audi partnered with Cognata to create virtual cities and simulate real-world conditions for their simulations.
V2x — This segment is in its very early stages. Only a small set of corporates and startups are working on this technology and even fewer are starting to test it. Qualcomm is the only corporate that has invested heavily in this space, and it is working with startups like Savari to develop the necessary software and chipsets for vehicles and infrastructure.
Finally
As the seed stage funding continues to drop, we will see fewer new startups entering this space. I think there are gaps in the V2x and simulation segments for newcomers, but the barrier for entry high in almost all the other sectors of this industry.
This is a race towards autonomy, and whoever racks more real-world miles gets to Level 5. Substantial costs and challenges associated with developing, building, and training AVs, Incumbents are looking to collaborate rather than compete (mainly to offset financial burden and share resources). So, we might see more partnerships and M&As in the future.
