Seeing The Road Ahead: The Undervalued Self-Driving Asset of Data

This is the 4th in a series of notes based on our deep dive into computer perception for autonomous vehicles. Autonomy is a question of when? not if? In this series, we’ll outline our thoughts on the key components that will enable fully autonomous driving. See our previous notes (Computer Perception Outlook 2030What You Need To Know About LiDAR, The Importance of Cameras To Self-Driving Vehicles).

A self-driving car will only be as strong as the data it trains on.

The importance driving data will play in developing fully autonomous vehicles is often understated, and companies that possess large and high-quality driving data sets are much further ahead of their peers than some may think. Driving data is key because it is the core input that will train the artificial intelligence models that operate autonomous vehicles. The more data these models have, the more scenarios they can prepare for, and, in turn, the stronger the entire system becomes. However, obtaining good driving data is not an easy task, and it is virtually impossible to gather data on every single driving scenario in all types of weather conditions. Due to improvements in computer graphics technology, many are relying on simulated data to train their self-driving models.  In the paragraphs below, we dive deeper into the pros and cons of using simulated data versus real data and identify who are the data leaders among self-driving car companies.

Simulation vs. real data. We recently spoke to a computer vision expert at the University of Michigan about the difference between using simulated data versus real-world data. He believes that simulated training data is valuable because most of the data collected during on-road driving is innocuous. Real-time driving data is only very interesting when there is a critical event or an unusual scenario. Simulation data lets you test on critical events constantly. However, he also noted the AI in simulated data is still programmed by a human and those AI tend to act differently than humans on the road. All autonomous systems will train with some simulated data and will therefore require fine tuning to factor in the difference between human and machine driving styles. That said, we do not want to underestimate the value of real data and believe capturing non-programmed scenarios will play a key role in preparing AVs for all situations that may arise.

Waymo’s large data lead.  The more miles an autonomous vehicle drives, the more real data the system can capture, the more robust the system can become. Companies that are approved to test autonomous driving in California are responsible for recording and publishing the number of autonomous miles driven. As of April 1st, 52 companies have been issued permits to test autonomous vehicles in California, and as shown in the graph below, Waymo has driven 352,545 as of November 30th, 2017. As of February 2018, Waymo had announced they have driven over 5 million miles in total. This announcement came only ~3 months after they announced crossing the 4-million-mile mark. While testing takes place in many states other than California, these data points suggest that Waymo has a very large data lead over their peers, which may translate to a large lead in the race to full autonomy.

Tesla lurking in the shadows. While Tesla is one of the 52 companies approved to test autonomous vehicles in California, Tesla did not test on state roads in 2017. However, the company acknowledged in the report they filed with California DMV that Tesla conducts testing to develop autonomous vehicles via simulation, in laboratories, on test tracks, and on public roads in various locations around the world. Tesla also highlighted that they have a fleet of hundreds of thousands of customer-owned vehicles that test autonomous technology in “shadow-mode” during their normal operation. Shadow mode is a feature that runs in the background without actuating vehicle controls in order to provide data on how the features would perform in real-world and real-time conditions. This has allowed Tesla to gather billions of miles of passive real-world driving data to develop its autonomous technology. This data is extremely valuable in training autonomous vehicles to interact with the real world, and, in our eyes, makes Tesla one of the top contenders in the race for full autonomy.

Disclaimer: We actively write about the themes in which we invest: virtual reality, augmented reality, artificial intelligence, and robotics. From time to time, we will write about companies that are in our portfolio.  Content on this site including opinions on specific themes in technology, market estimates, and estimates and commentary regarding publicly traded or private companies is not intended for use in making investment decisions. We hold no obligation to update any of our projections. We express no warranties about any estimates or opinions we make.

Seeing the Road Ahead: The Importance of Cameras to Self-Driving Vehicles

This is the third in a series of notes based on our deep dive into computer perception for autonomous vehicles. Autonomy is a question of when? not if? In this series, we’ll outline our thoughts on the key components that will enable fully autonomous driving. See our previous notes (Computer Perception Outlook 2030, What You Need To Know About LiDAR).

If a human can drive a car based on vision alone, why can’t a computer?

This is the core philosophy companies such as Tesla believe and practice with their self-driving vehicle initiatives. While we believe Tesla can develop autonomous cars that “resemble human driving” primarily driven by cameras, the goal is to create a system that far exceeds human capability. For that reason, we believe more data is better, and cars will need advanced computer perception technologies such as RADAR and LiDAR to achieve a level of driving far superior than humans. However, since cameras are the only sensor technology that can capture texture, color and contrast information, they will play a key role in reaching level 4 and 5 autonomy and in-turn represent a large market opportunity.

Mono vs Stereo Cameras. Today, OEMs are testing both mono and stereo cameras. Due to their low-price point and lower computational requirements, mono-cameras are currently the primary computer vision solution for advanced driver assistance systems (ADAS). Mono-cameras can do many things reasonably well, such as identifying lanes, pedestrians, traffic signs, and other vehicles in the path of the car, all with good accuracy. The monocular system is less reliable in calculating the 3D view of the world. While stereo cameras can receive the world in 3D and provide an element of depth perception due to dual-lenses, the use of stereo cameras in autonomous vehicles could face challenges because it’s computationally difficult to find correspondences between the two images. This is where LiDAR and RADAR have an edge over cameras, and will be used for depth perception applications and creating 3D models of the car’s surroundings.

Camera Applications. We anticipate Level 2/3 and Level 4/5 autonomous passenger cars will be equipped with 6 – 8 and 10 – 12 cameras, respectively; most will be mono-cameras. These cameras will play a prominent role in providing nearly 360-degree perception and performing applications such as lane departure detection, traffic signal recognition, and park assistance.

$19B Market Opportunity. Given cameras are the primary computer perception solution for advanced driver assistance systems (ADAS), the camera market is currently the largest computer perception segment and represented a $2.3B opportunity in 2017. While growing adoption of ADAS enabled cars will continue to act a near-term catalyst, adoption of fully autonomous vehicles (Level 4/5) equipped with 10+ units per car, will be the tailwind taking this market to $19B by 2030 (18% CAGR 2017 – 2030). As displayed in the chart below, the bulk of sales will be in the form of mono-camera systems. Note our forecast is driven by our 2040 Auto Outlook and only includes passenger vehicles. Fully autonomous heavy-duty trucks will also leverage similar computer vision technology, and when factoring these sales, the total camera market could be 1.5x- 2x larger.

Disclaimer: We actively write about the themes in which we invest: virtual reality, augmented reality, artificial intelligence, and robotics. From time to time, we will write about companies that are in our portfolio.  Content on this site including opinions on specific themes in technology, market estimates, and estimates and commentary regarding publicly traded or private companies is not intended for use in making investment decisions. We hold no obligation to update any of our projections. We express no warranties about any estimates or opinions we make.

Nvidia; Buy More GPUs, Save More Money

  • NVDA shares dropped 8% today on the announcement they temporary stopped autonomous testing and market sell off.
  • While the company did not comment on timing, we expect testing to resume in the next 3 months.
  • CEO Jensen Huang held a keynote today. The message was more GPU’s save money.
  • The company announced 5 new products targeted at AI, autonomy, and VR.

Nvidia started its GPU Technology Conference on Monday with their focus on AI & Deep Learning. On Tuesday, CEO Jensen Huang gave a keynote on Nvidia’s latest developments in a few key product categories, but Nvidia’s stock dropped by 8%. We attribute 2/3 of the decline to the company announcing they’ve temporary stopped autonomous vehicle testing until they receive a diagnostic report from the Uber accident last week, and 1/3 from the broader tech sell off.

Buy more GPUs, save more money. The message from Jensen Huang was if you buy Nvidia’s GPUs, you can save money. The idea is using an Nvidia architecture requires less hardware that consumes less energy.  That said, users of these systems are solving more advanced problems like AI, which will require an increase in total spending. We remain comfortable with our Nvidia revenue growth estimates of 31% in CY18 and 21% in CY19.

New Products. Nvidia made product announcements today:

  • RTX (AI and blockchain). On the workstation front, the company announced the Quadro GV100 with (real-time ray tracing) RTX Technology. The RTX technology offers a measurable improvement in rendering times.
  • DGX-2 (AI and blockchain). Nvidia also announced the DGX-2 platform for AI. The DGX-2 claims to be the first to offer 2 Petaflop single server deep learning system. Nvidia compared this platform to a traditional hyperscale center, which would have 300 dual-core CPUs and cost about $3M. The DGX-2 is may seem expensive at $399K, but it takes up 1/60th of the space and consumers 1/18th of the power.
  • Isaac SDK (AI) Adding artificial intelligence to robots for perception, navigation, and manipulation. Nvidia showed a video of one of its first Isaac projects, a two-wheeled delivery robot named Carter.
  • Drive Orin (Autonomy). Drive Orin is comparable to two Drive Pegasus supercomputers while being smaller in physical size. Jensen shared that many customers were including two Drive Pegasus chips in each vehicle, and it made sense to package the same computing power into one product.
  • Drive Constellation (Autonomy). This is a datacenter solution used to test and validate self-driving vehicles in virtual reality.
  • Project Wakanda (VR). See details below.

Nvidia committed to an autonomous future. Given the Tempe accident, Jensen spent most of the self-driving segment of the keynote talking about the importance of safety, and why fully-autonomous future means safer transportation. He also reiterated this his belief that self-driving cars are “probably the hardest computing problem that the world has ever encountered”.  The company did not give a timeline of when testing will resume, but given the hold is based on the Uber review, we would expect it to take a few months. Autonomous simulation will play an increasingly important role in the future. Drive Constellation can run thousands of virtual worlds, each while running thousands of scenarios in order to collect more data. For example, 10,000 constellations can simulate about 3 billion miles in a year, significantly more than 5-10 millions driven each year by the current fleets of test vehicles.

Project Wakanda hints at the future of relationship between man and machine. Nvidia closed the keynote by unveiling what has been dubbed “Project Wakanda.” Similar to what we saw in Black Panther, Jensen showed a driver operating a vehicle in virtual reality, while sitting in a fully-equipped cockpit. Next, a third screen appeared and showed a driverless car at a remote location.

While Nvidia didn’t share much about its direction with this project, Jensen did share that he views virtual reality as a way to provide humans with teleportation – augmented with autonomous machines.

Disclaimer: We actively write about the themes in which we invest: artificial intelligence, robotics, virtual reality, and augmented reality. From time to time, we will write about companies that are in our portfolio. Content on this site including opinions on specific themes in technology, market estimates, and estimates and commentary regarding publicly traded or private companies is not intended for use in making investment decisions. We hold no obligation to update any of our projections. We express no warranties about any estimates or opinions we make.

Seeing the Road Ahead: What You Need to Know About LiDAR

This is the second in a series of notes based on our deep dive into computer perception for autonomous vehicles. Autonomy is a question of when? not if? In this series, we’ll outline our thoughts on the key components that will enable fully autonomous driving. See our previous note (Computer Perception Outlook 2030) here.

LiDAR sensors will play a prominent role in the road to full autonomy. Although we think all computer perception technologies, such as cameras and RADARs, are important, the integration of LiDAR provides a sensor suite with increased performance and redundancy that is not attainable with cameras and RADAR alone. Most importantly, it provides another layer of data beyond what humans can perceive. In hopes of building a system that is safer and more efficient than humans, more data is always better. LiDAR’s primary applications play an important role by supplementing the data gathered by cameras and RADARs with object recognition, distance estimation, and advanced mapping capabilities. Given how important LiDAR will be to this emerging theme, we believe the LiDAR market opportunity solely for autonomous driving can grow to $16B by 2030, as higher volumes of level 4/5 self-driving vehicles enter the roads. While we do not believe the real inflection will begin until 2020, key players are emerging today and we believe there are 3 key points everyone interested in this theme needs to know today.

1) LiDAR comes in different forms. The first thing one should know about LiDAR is there are multiple kinds of LiDAR technologies. As displayed in the exhibit below, each type of LiDAR offers advantages and disadvantages to price, performance, and reliability, but it’s important to know that the industry is moving from Mechanical LiDAR to Solid State LiDAR due to lower costs and higher reliability. However, to make it slightly more complicated, there are multiple kinds of Solid State LiDAR, and once again, each is better utilized in certain applications. Today, most OEMs are still testing multiple LiDAR technologies, and the industry is still looking for commercialized LiDAR technologies that meet safety and performance requirements at mass-market production levels.

2) Key components and who makes them. Regardless of the type of LiDAR technology, almost all systems rely on similar core components. As displayed in the interactive exhibit below, these key components include laser emitters, photodetectors and advanced integrated circuits (ICs) to process the high-density LiDAR information. Historically and still largely today, Tier 1 auto suppliers such as Bosch, Denso, and Delphi are the primary manufactures to most components and systems going into automobiles. However, due to the complexities of manufacturing LiDAR systems, auto OEMs will need to work with new kinds of suppliers. This will include buying directly from LiDAR sensor manufacturers like Velodyne and Quanergy, as well as optical component players such as Daktronics, AMS AG, Fabrinet, Lumentum, Finisar, II-VI, and Viavi. Use the grey toggles below to dive deeper into these key components, as well as identify the leading component manufacturers.

Source: Loup Ventures

3) $16B market opportunity. Although we believe the inflection point for LiDAR systems will not occur until 2020, we anticipate this market can quickly grow from <$100M today to a $16B market opportunity by 2030 as a higher volume of Level 4/5 autonomous cars enters the road. (See computer perception model here.) Initially, demand will be driven by Mechanical State liDAR, but as Solid State LiDAR manufacturers are able to introduce lower price point systems at scale, we expect these technologies to take over the market. Specifically, we believe MEMs-based scanning LiDAR will be the primary solution for meeting the high volumes and LiDAR performance requirements needed for autonomous vehicle deployments as early as 2020. However, once pure Solid State LiDAR (2D/3D Flash, Optical Array Phased, FMCW) are ready, MEMs technology will lose share to these lower cost and higher reliability technologies.

Disclaimer: We actively write about the themes in which we invest: virtual reality, augmented reality, artificial intelligence, and robotics. From time to time, we will write about companies that are in our portfolio.  Content on this site including opinions on specific themes in technology, market estimates, and estimates and commentary regarding publicly traded or private companies is not intended for use in making investment decisions. We hold no obligation to update any of our projections. We express no warranties about any estimates or opinions we make.

Tracking Progress in Electric & Autonomous Vehicles

We were set to publish a note on targets from major automakers to ramp production of electric and autonomous vehicles when we heard the news from Tempe.  Regardless of who was responsible, it’s a tragic reminder about the dangers of road safety and reminds us that the move to self-driving cars will take longer than most think. We’ve been asking ourselves: what’s the right way to think about the risks of autonomy? Is the threshold for AVs zero accidents, or is it an improvement relative to human drivers? In the end, we continue to believe humans should not drive, and that traffic and pedestrian fatalities will decline under autonomy.

Tesla’s activity in electric and autonomous vehicles (EV and AV) is well documented. The electric car maker’s unlikely rise as a leader in EV and AV has forced other manufacturers to turn their attention to emerging technologies. Fuel cell technology had been considered the alternative fuel of the future, but Tesla’s success has altered the mindset of the industry and pushed other carmakers to develop battery-powered vehicles. Separately, the development of autonomous vehicles is a hyper-competitive space, with companies pouring billions of dollars in R&D for self-driving technology. Tesla’s Autopilot is probably the most advanced AV system on the road today, but its competitors are nipping at their heels. It’s clear that autonomous vehicles are the future, what’s much murkier is who the winners are going to be and when they are actually going to win.

Toyota

  • EV: In December of 2017 Toyota announced they were ramping up production of various forms of electrified vehicles (i.e. hybrids, battery-powered, fuel cell). They plan to offer an electrified version of every model they have in production. By 2030 they hope to have sell 5.5m electrified vehicles, with 1m of those being zero-emission (i.e. fully electric). Initially, they will roll out their EVs in China before moving to other markets like India, Europe, and the U.S.
  • AV: The Toyota Research Institute is Toyota’s R&D arm that has been doing most of the work in autonomy for Toyota. At CES 2018, they showcased the third iteration of their self-driving vehicle, Platform 3.0. Its major improvement was the 4 long-range LiDARs on the roof that allow it to “see” 200m out all 360 degrees around the vehicle. Toyota says this makes it one of the most perceptive AVs being tested today. In late March 2018, Toyota was rumored to be in talks to have Uber supply them with their self-driving technology, which way mean Toyota’s technology isn’t advancing as fast as they would have hoped.

Volkswagen Group

  • EV: This past week VW came with some big news for EV plans: $25m in battery supplies to begin producing 3m vehicles a year in 16 factories by 2025. Volkswagen expects to spend up to 50B Euro on batteries to supply the initiative. The battery push is a continuation of VW’s aggressive electrification strategy. In September it was announced the they would be producing electric versions of all 300 vehicle models across the group’s 12 brands, a $24B investment.
  • AV: Volkswagen is also heavily involved in AV, even showcasing a concept AV at last month’s Geneva Motor Show without a steering wheel or pedals. Beyond that, they have teamed up with AV software rockstars Aurora and announced production plans for a 4-seat fully autonomous bus.

Ford

  • EV: Ford has been vocal about their commitment to electrifying their vehicles, planning to invest $11 billion in by 2022 and have 40 hybrid and fully electric models (upping their previous announcement of investing $4.5B by 2020).
  • AV: Ford revealed its self-driving “platform” at CES in January, indicating Ford is more interested in being the OS of mobility’s future than simply a carmaker. The company also will invest $1b over the next 4-5 years in Argo AI, who will develop the autonomous technology for Ford to use with their vehicles. Ultimately, Ford hopes to operate a ride-hailing service with an autonomous fleet of Ford cars. No timetable has been set.

Honda

  • EV: Honda got into the electrification game in October of 2016 when they set up an electric vehicle division in their R&D department. They were previously focused on hydrogen fuel cell technology for an alternative fuel system. Then in summer of 2017, they pledged to electrify two-thirds of their globally-sold cars by 2030. Currently, about 5% of sales are EVs.
  • AV: Honda has stated they plan to come out with level 4 autonomous vehicles, meaning they can navigate highways and city roads in most conditions, by 2025. By 2020 they hope to have level 4 cars on the road. It was reported they were in talks to form a partnership with Alphabet-owned Waymo, but those discussions appear to have fallen through.

Renault-Nissan

  • EV: Nissan was an early adopter of electric vehicles with fully electric Leaf in 2010, which has gone on to become the world’s best-selling electric vehicle with over 300k sold. Since then Nissan, French carmaker Renault, and Mitsubishi have formed a strategic partnership and announced their ‘Alliance 2022’ plan in September of 2017. The plan endeavors to launch 12 new zero-emission vehicles by 2022. Furthermore, Infiniti (owned by Nissan) announced in January that they would be moving to an all-electric lineup by 2021.
  • AV: In the Alliance 2022 plan, the three partners state they will introduce 40 vehicles with differing levels of autonomy (i.e. driver assistance up to full autonomy). Reading between the lines means they anticipate having full autonomy (level 4/5) figured out in less than 5 years. Nissan also unveiled a plan to launch a self-driving taxi service in Japan, called Easy Ride. Initially only offering short rides between Nissan HQ and a popular shopping mall just under 3 miles away, Nissan and their partner DeNa said they will expand it to a much wider market by 2020.

Hyundai

  • EV: Hyundai was slow to adopt electrification, finally announcing in 2017 that they would add 38 green cars in the next 5 years. They will not all be battery-powered, some will be plug-in hybrids and some will be hydrogen-powered fuel cell vehicles, which Hyundai still considers to be the long-term future of zero-emission vehicles. They have differentiated themselves from competitors by focusing on range and are planning to offer two vehicles at the high-end of EV’s range of around 300 miles.
  • AV: Hyundai, like VW, has partnered with Aurora to develop AV technology and manufacture autonomy-capable models to sell to fleet services. They anticipate having AVs on the market in 2021, most likely at level 4. Beyond the Aurora partnership, Hyundai had AVs shuttle athletes and attendees to and fro at the 2018 Winter Olympics and had showcased their concept for a mass-market AV back in 2016.

General Motors

  • EV: GM was also an early player in EVs, releasing the Chevrolet Volt plug-in hybrid in 2011, but they have strengthened their commitment to the space and are striving to bring 20 new EVs to market by 2023. They also are aiming to sell 1m EVs a year by 2026. The competitive advantage they are seeking and touting is profitability through leveraging their superior scale in manufacturing and capital.
  • AVIn one of the most ambitious AV announcements to date, GM announced they would be making an AV without a steering wheel or pedals by 2019. They also just announced they will be investing $100m in the production of the car, dubbed the Cruise AV (and based on the Volt), in two U.S. factories. They acquired Cruise Automation, an AV startup, in March 2016 for an undisclosed amount. GM wants to be a full stack AV company where they control all aspects of the production and technology. They also have announced plans to launch to their own AV ride-hailing fleet in 2019.

Fiat Chrysler

  • EV: At the 2018 Geneva Motor Show FCA (Fiat Chrysler Automobiles) CEO Sergio Marchionne admitted that his company made a mistake in underestimating the rise of electric vehicles. Fortunately for FCA, this admission comes after they have implemented plans to electrify their brands’ cars, though only one vehicle is rumored to be all-electric – a high-end Maserati – and the rest will be hybrids. Bottom line is FCA is a laggard in adopting EV technology.
  • AV: Though they may be behind in EVs, FCA has been heavily involved in the development of AVs. They have supplied Waymo with hundreds of minivans for testing and reached a deal this year to supply Waymo with “thousands” of minivans for use in Waymo’s new ride-hailing fleet. FCA also joined BMW and Intel/Mobileye in an alliance to develop AVs. FCA’s role in the partnership will be to add resources and expertise as well as act as the vendor for the planned AV platform.

Volvo

  • EV: Last year Volvo pledged to phase out ICE engines and by 2019 only offer hybrid or battery-powered vehicles. It was a major announcement being one of the first automakers to commit fully to EVs. From 2019 to 2021 they plan to introduce 5 all-electric vehicles and recently unveiled a high-end model from one of their sub-brands, Polestar, that will challenge Tesla’s Model S.
  • AV: Volvo began a unique program called Drive Me where they were going to give 100 AVs to families in Sweden for testing, but in late 2017 scaled it back as the technology wasn’t ready yet. Separately, Uber announced last November that they would be buying as many as 24,000 self-driving Volvos, once the tech is viable, to put in its ride-hailing network. The deal’s terms were undisclosed but estimated to be worth around $1B. They had previously made a $300m deal together to develop AV technology.

Bottom line. We believe that autonomous drivers will be many times safer than human ones but will take longer than most people think to develop. Here at Loup Ventures the initiatives and timelines for AVs given by these companies strike us as ambitious and unlikely to be met. It is unlikely we see something like a fully operational autonomous fleet shuttling folks around in the next 5 years. We are enthusiastic and optimistic about this future of mobility but are managing expectations for how long it will truly take for the ubiquitous adoption of autonomous vehicles. As for electrification, we believe there is enough momentum and fewer hurdles for adoption that widespread use of battery-powered vehicles is imminent.

Disclaimer: We actively write about the themes in which we invest: virtual reality, augmented reality, artificial intelligence, and robotics. From time to time, we will write about companies that are in our portfolio.  Content on this site including opinions on specific themes in technology, market estimates, and estimates and commentary regarding publicly traded or private companies is not intended for use in making investment decisions. We hold no obligation to update any of our projections. We express no warranties about any estimates or opinions we make.