SoftBank Founder Masayoshi Son’s Self-driving Kingdom



2018-08-14 08:44   Leiphone

Bill Gates commented that Masayoshi Son is an adventurer like him. Since SoftBank was founded in 1994, Masayoshi Son has built up an information technology empire in just 24 years. “If someone makes business choice based on a thick pile of documents and if he always looks into something happens decades or even 300 years later from now, it will be not easy to make such a person fail completely.” Masayoshi Son is this kind of person. 

The internet is still booming, while Masayoshi Son has already taken aim at the era of Internet of Things (IOT) decades later. Robotics, IOT and Artificial Intelligence (AI) are the fields that he wants to invest into in the future. He once described that “by 2040, everyone will carry an average of more than 1,000 networking devices with them.” At that time, mobile phones, cars and daily necessities will all be connected with the devices, generating a lot of data. 

 “The era is so great that I feel sleeping is a waste of time.” Mr. Son expressed his desire for the future and passion to embrace the era of Internet of Things. Actually, he has already increased his investment in AI and autonomous driving. Now his “Autonomous Driving Empire” is quietly emerging.  

Starting up business in autonomous driving is difficult, but investing is even more difficult. No investors want to miss the market share but they are afraid that one careless move may result in losing the whole game. Masayoshi Son is leading his well-funded Vision Fund, aiming at unicorns from upstream and downstream of the autonomous driving industry chain and building up an Autonomous Driving Empire.

Masayoshi Son’s giant fund——Vision Fund

Sensors are playing an increasingly more important role in autonomous vehicles as eyes. Self-driving companies are also slightly different in their choice of sensor solutions. Automakers represented by Tesla choose the more cost-effective combination of “millimeter wave radar + camera”; Waymo and Baidu use a high-cost lidar that is slightly better in accuracy. Apple, Uber and Roadstar generally use a multi-sensor fusion solution which is currently widely believed by the industry to be the future trend.

Vision Fund has layouts in both camera and lidar, aiming at the potential enterprise Nauto. Light, which received investment from Vision Fund a few days ago, is a veteran player in the area of precision cameras and Son plans to apply its technology to autonomous vehicles. Innoviz, the dark horse in the area of solid-state lidar, is also fancied by SoftBank.

Camera startup Nauto

Investors such as Vision Fund’s co-venture firm Greylock Partners made a $159 million investment in Nauto last July, a startup whose previous investors included BMW, Toyota, GM and insurance firm Allianz.

There must be something unusual about Nauto that attracts investment from SoftBank and many well-known automobile companies.

 “Making a regular car into a Tesla” is exactly what Nauto is doing. Whether it's the self-driving test cars coming and going in California's Bay Area or the Tesla cars equipped with Autopilot, accumulating massive mileage data becomes the only way for cars to achieve autonomous driving. The main business of Nauto is to collect autonomous driving data and develop relevant software. Its rear camera products can convert the driving behavior of traditional car into more valuable road data.

Nauto detects the drivers’ eyes to see if s/he is looking at the cell phone or dozing via a rear loading device (a recorder loaded with advanced computer vision processing software developed by Nauto) that can collect and process visual data, adding the monitoring of head and upper body positions to improve the accuracy of identification. The data analysis is used to judge whether the driver has dangerous driving behavior and to provide safe driving advice. The device also has a camera facing away from the car, which is used to take pictures of the road and integrate the traffic and driver data to record the overall data of how the driver controls the car in different road conditions. The complete systematic data of driver, vehicle and road are eagerly desired by many self-driving players. 

Nauto’s deep learning aftermarket products are very attractive to some business fleets and insurance companies. Nauto devices have been applied to the San Francisco Citywide Taxi fleets. The small camera in the device can detect and make alerts to the drivers’ behaviors such as drunk driving and cell phone use, which may also help auto insurance companies assess risk, prevent fraud and reward high-quality customers. In addition to cooperation with business fleets and insurance companies, Nauto has also taken advantage of its autonomous driving data collection to work closely with BMW and Toyota on developing autonomous vehicles.

Precision Camera Startup Light

Vision Fund’s investment in Light is interesting. Dave Grannan, co-founder and CEO of this video startup, won $121 million investment from Vision Fund after three meetings with Masayoshi Son. Light is known for its R&D of “information dense” camera, but it’s hard for many people to imagine what it has to do with autonomous vehicles. Son sees Light as a potential alternative for lidar. 

Although the Vision Fund is ostensibly investing in a precision camera startup, in fact it is Son’s ambition of laying out autonomous driving. 

The camera serves as the eyes of the autonomous vehicle, and Son told Grannan that the Light camera will create the greatest value after application in the autonomous vehicle. That’s a wake-up call for Grannan which also means that Light is about to get into autonomous driving. Light can use complex algorithms to compose the images collected by different camera modules into a high-quality image that contains abundant in-depth information. What’s more, the camera is lighter and cheaper.

Solid State Lidar Startup Innoviz

A light-small and low-cost solid state lidar is considered one of the sensors essential to the mass production of self-driving cars. SoftBank certainly will not miss it. In the solid-state lidar battlefield, there are now dozens of domestic and foreign lidar companies, from early entrants Velodyne to Quanergy and Luminar, etc. The Israeli startup Innoviz Technologies, founded in 2016, is a dark horse, and SoftBank led its Series B round funding. 

Just two years after its founding, Innoviz has signed two big deals: one is the solid state lidar supply for BMW's upcoming self-driving cars and the other is partnership with a domestic auto parts supplier, Jinwei Hengrun, to promote products and technology into the Chinese market.

The high-definition solid-state lidar (hd-ssl) is the first product launched by Innoviz, costing less than $100 and is no more than 5cm in length, width and height. Last May, InnovizPro, a mass-produced version for the aftermarket (supplied to automaker for self-driving R&D and testing), was launched with a detecting range of 150 meters and a dimension of 18cm*8cm*8cm. For the OEM market, the mass-production version of InnovizOne meets the requirements of vehicle specifications. It can perform 3D scanning in high definition, with a resolution of 6 million pixels per second and a depth measurement accuracy less than 2 cm, with a detection range of 200 meters and a dimension of 5cm*5cm*5cm. InnovizOne's components are already in production, planning for shipment in 2019.  

HD Map

HD map is one of the indispensable conditions of autonomous vehicles. By then, the map will no longer be a purely vehicle-mounted system but will be used to guide the directions of unmanned vehicle. Vision Fund has also laid out in HD map, targeting Mapbox, a mapping service company that is dedicated to provide developers with development tools and open platforms.

Mapbox, HD Map 

On October 2017, Vision Fund led the investment of $164 million in Mapbox, an American mapping service provider. It was reported by foreign media that SoftBank has high hopes for the company as an important mapping software for autonomous vehicle development. With this round of financing, Mapbox will make greater investment in in-car navigation and autonomous driving.

Mapbox started from making web maps, focusing on open source applications. Now the platform has developers from over 200 countries, and the registered developers have exceeded one million, which is very important for a technical platform.

The Mapbox Drive (SDK) was launched and installed into cars with corresponding sensors through crowdsourcing mode and cooperation with automakers. These cars will upload location information to map the world. After the SoftBank investment, Mapbox has also been working on the R&D of AR and VR maps. Recently, Mapbox launched a new SDK, which allows developers to complete AR maps quickly. This technology will be directly applied to self-driving cars.. In addition, Tesla also chose Mapbox to help it make digital maps

Self-driving Chips

SoftBank bought all shares of ARM Holdings for $32.4 billion in 2016, a UK semiconductor intellectual property provider which shipped 15 billion chips in 2015.  In the following March, SoftBank sold its 24.99% stake of ARM to Vision Fund for $8.2 billion. Soon after, chip giant Nvidia was acquired by SoftBank as well, which bought $4 billion stake in Nvidia with a 4.9% share. A few months later, SoftBank transferred its 4.9% stake to Vision Fund, which also became the fourth largest shareholder in this graphics chipmaker.

Chip Giant Nvidia

Nvidia’s entry points in the autonomous driving area are mainly in AI and machine learning. Its GPU has few competitors in the area of general computing and is the industry's preferred choice in supporting CNN algorithms

Drive, “the end-to-end platform” that Huang Renxun says has revolutionized the self-driving industry, has taken shape. This is a full self-driving overall platform solution, including ASIL-D OS(Drive OS), deep learning (Drive PX), computer vision SDK (Driveworks SDK) and self-driving app Drive AV.

Drive PX iterated to develop computers for different levels of autonomous vehicles, including Drive PX 2, Drive PX Xavier and DRIVE PX Pegasus. Drive PX series are specially designed for autonomous driving, combining technologies such as deep learning, sensors fusion and surround vision, etc. The self-driving software stack developed based on Drive PX is capable of understanding in real-time what’s happening around the vehicle, precisely locating itself on an HD map and planning a safe path forward.

Drive PX: The first supercomputers used in self-driving tests

Drive PX 2: Single processor performance enables Level 2 to Level 3 autonomous driving, and parallel use of multiple Drive PX 2 enables Level 4 to Level 5 autonomous driving

Based on DRIVE PX2, Nvidia has launched mass production version of DRIVE PX2 for AutoCruise and DRIVE PX2 for AutoChauffeur. The former is mainly used for high-speed autonomous driving and HD mapping, while the latter is used for point-to-point autonomous driving in specific scenes. It’s worth mentioning that, Tesla also uses a customized version of Drive PX2 in its mass-produced model.

Drive PX XavierA single processor enables Level 4 to Level 5 autonomous driving. It’s the world's first autonomous machine processor whose “complex system on chip(SOC)” is available for simulation, training and autonomous driving. Xavier samples will be available to some partners in the first quarter and be launched into market in the fourth quarter of 2018.

DRIVE PX PegasusA mass-produced supercomputer specially developed for L5 self-driving taxi. Pegasus operates at 320 teraflops, consuming 500 Watts, more than 10x the performance but also power consumption of Xavier with 30 teraflops and 30 Watts.

Nvidia is a “great partner” of self-driving enterprises, supported by automakers such as Tesla, Toyota, Audi and Volvo as well as tier 1 suppliers such as Bosch and ZF. Currently its circle of friends has expanded to more than 300 companies, not all of which are automakers, but also map companies and lidar developers. More than 100 startups have developed self-driving software stacks by Drive PX platform. 

Semiconductor Giant ARM

ARM is a chipmaker that does not produce chips. After 27 years’ development, ARM has gained some share in the mobile chip market and even Apple and Samsung are using its chip design solutions. But ARM does not want to squander the opportunity to get ahead of the market in the new era of autonomous driving.

Cortex-ACortex-R and Cortex-M series MCU are three major product lines of ARM. In addition, ARM Mali series GPU and Project Trillium also support autonomous vehicles. 

Cortex-A is a high-performance computing processor, supporting ADAS, partial center console, GPS and DCM (Data Communication Module). The first generation of DynamIQ processors released last year, cortex-A75 and cortex-A55, have improved the performance under two different power budgets to a whole new level, enabling the self-driving technology to achieve a safer and faster response. 

Cortex-R is embedded and used in the products where high stability is required, such as EPS, ABS, PTO and battery management. When it was officially owned by SoftBank, ARM launched cortex-R52 processor for autonomous driving. Based on the brand-new ARMv8-R architecture, cortex-R52 enables self-driving cars powerful ECU computing power and more complex software stack for security management, which can guarantee that autonomous vehicles will become more secure in overtaking, accelerating and braking. ST Microelectronics is ready to use it to build highly integrated system on chip (SoC) for application in the automotive market 

Cortex-M is a small low-power processor supporting devices such as airbag, GPS and EPS.

Mali series GPU is specially designed for auto images processing. ARM first announced ISP (Image Signal Processor) — Mali-C71 processor for autonomous driving in April 2017, which can process images quickly under extreme conditions, meeting the automobile requirement. 

ARM announced a new machine learning (ML) compute platform—Project Trillium in the beginning of this year. Project Trillium is a suite of Arm IP including highly scalable processors that will deliver enhanced machine learning (ML) and neural network (NN) functionality, which will be applied in autonomous driving in the future. ARM also announced the introduction of the first common industry framework for building an economic and scalable common security framework, called Platform Security Architecture (PSA). To provide better solutions for M2M communication of autonomous vehicles, ARM integrated SIM cards into the processor by Kigen SIM technology which enables drivers to gain real-time information from vehicle sensors. 

Vision Fund made relentless investment in the downstream (actual model) technology companies, with the $2.25 billion investment in GM Cruise causing a stir this year. 


Vision Fund made $2.25 billion investment in GM Cruise, a competent player in autonomous driving area in May. SoftBank planned to invest the $2.25 billion in GM Cruise in two tranches, with the first $900 million to be paid when the deal is completed and the remaining $1.35 billion to be paid when Cruise Automation's autonomous vehicles are truly commercial. Vision Fund will then take a 19.6% stake in GM Cruise whose value will increase to $11.5 billion. 

Though GM Cruise is not as high-profile as Waymo, Uber and Baidu, it actually develops fast and best in the autonomous driving area and was at the TOP of the 2017 Self-driving Competitiveness Rankings released by Navigant Research, an American authoritative market research institution. The 2017 Latest Self-driving Testing Report announced by DMV showed that Cruise’s road mileage reached 125,000 miles last year and the frequency of human intervention was reduced to 105/12500=0.84 times/thousand miles, a marked improvement over last year’s 18.5 times/thousand miles. It was second only to Waymo. 

Currently, Level2 super Cruise smart driving technology has been applied in Cadillac CT6 which was also the first autonomous driving system enabling fully hands-free on highways. With a lot of road test data, GM Cruise also began to plan mass production of autonomous vehicles. The new generation Cruise autonomous vehicles cancelled steering wheels and brake, which reached Level4. GM Cruise was obviously ahead of others in the preparation of mass production and Navigant Research also expressed that this automobile giant “is very likely to succeed in highly self-driving layouts in the coming years.”

Brain Corporation

Brain Corporation, which develops robotic autonomous navigation technology, has also received $141 million in investments from Vision Fund and Qualcomm. The main business of Brain Corporation is controlling the robot to navigate using visual cues and landmarks, allowing it to see and avoid people and obstacles in a way which is similar to the “last mile of a self-driving car.”

Overview of Masayoshi Son’s “Investment Chessboard”

In the overview of the investment layout of Vision Fund, it is either upstream or downstream, or key technologies from the perspective of the invested enterprises. As Lydia, a SoftBank investor, says, “we do not invest all leading companies of the area, from TOP1 to TOP4.” Vision Fund chases specifically after the potential ones in the area and many of its rivals are also surprised by its unique investment approach in technology. 

In addition to being an investor in unicorns in the industry, SoftBank is also cultivating its own autonomous driving team internally. SoftBank Group and Advanced Smart Mobility established a joint venture, SB Drive in 2016, making R&Dt of autonomous buses. In July, SB Drive and Baidu Japan Inc. reached cooperation. Early next year, 10 Apolone Mini electric buses jointly developed by Baidu and Jinlong United Automobile will be shipped to Japan. SB Drive also plans to launch the commercial use of autonomous buses in Haneda Airport in 2020. 

Vision Fund continues its investment in autonomous driving.  Vision Fund may invest $1billion in Sense Time, an AI company, Bloomberg News said, citing people familiar with the situation. But the deal has yet to be finalized and the amount and manner of transaction are likely to change. However, it’s undeniable that Sense Time Technology, which is dedicated to computer vision and deep learning, is looking to get into the autonomous driving field.

SoftBank’s autonomous driving layout is becoming clearer with key investments and independent R&D. With the advent of autonomous driving, Mr. Son’s “Self- driving Empire” is on the rise.

Mr. Son has made long-term layout for future Internet of Things. The acquisition of ARM symbolized the transformation of SoftBank from traditional mobile/internet industry to IOT which was reflected in Vision Fund, per Masayoshi Son during SoftBank’s early acquisition of ARM. Today, the Vision Fund’s investments cover many cutting-edge technologies such as autonomous driving, VR, ridesharing and IOT. It’s worth noting that, Vision Fund is different from conventional internal venture-capital departments, outside investors such as Public Investment Fund, the Vision Fund were asked to contribute a big slug of debt in the form of preferred units along with their equity checks. Of the nearly $100 billion that Vision Fund has raised, about $44 billion is in the form of preferred units, which SoftBank promises to pay an annual coupon of 7% over the fund’s 12-year life cycle. It means that SoftBank needs to pay about $3billion interest to creditors every year which is a huge sum. 

Vision Fund seems to be a huge bet of Masayoshi Son. He is confident that the next generation of information revolution is around the corner and IOT will be the future. Just as a decade ago, Masayoshi Son believed that the future of, an unknown company then, will be doomed to be different. 

 “The Vision Fund was just the first step, $100billion is simply not enough," CEO Masayoshi Son said. SoftBank will briskly expand the scale to $880billion investment in tech in the next years through iterations of its Vision Fund. Vision Funds 2, 3 and 4 will be established every two to three years. Recently, it’s reported by foreign media that SoftBank is in talks to raise a second tech investment fund, bigger than the previous $93 billion Vision Fund.

When some people joke that he spends too much money, the 60-year-old entrepreneur just smiled and said that, “It’s just the beginning now and we will continue investing at this rate for the next 5 years or even more.”