Software-Defined Vehicles a Forthcoming Industrial Evolution

Software-Defined Vehicles A Forthcoming Industrial Evolution

3 w

Software-Defined Vehicles—A Forthcoming Industrial Evolution | Contents

Contents

Introduction 1 1. Understanding "software-defined vehicles" 2 2. Driving forces of software-defined vehicles 3 2.1 Industrial development requirements 3 2.2 Consumer expectations 3 2.3 Value chain transfer 4 3. Gap between the Present and the Future 6 3.1 The automotive hardware and software architecture does not suit the software-defined vehicles 6 3.2 The traditional waterfall software development model has major limitations 11 3.3 Organizational structure and talent supply are major shortcomings of software-oriented automotive transformation 13 3.4 Obstacles from the supply chain system 14 4. What are the new opportunities 16 4.1 Industry value chain under the "software-defined vehicle" trend 16 4.2 Prominent position of software platforms 17 4.3 From Tier-2 to Tier-0.5 19 5. How should different enterprises respond 21 5.1 OEMs transform based on rational assessment and their capabilities 21 5.2 With "oppression from both ends", parts enterprises should seek self-transformation 23 5.3 Facing both opportunities and challenges, automotive software companies shall create value based on flexible positioning 25 Conclusion 26 Contact us 27 w

Software-Defined Vehicles—A Forthcoming Industrial Evolution | Introduction

Introduction

Over the past few years, standardized hardware of smartphones and computers gradually reached physical limits, pushing industrial transformation from hardware upgrades to software development. The automotive industry is different from the smartphone and computer industries in terms of hardware standardization and technology. Thus, the automotive industry is not ready to replicate the exact development pattern of the smartphone or computer. However, with increasingly standardized hardware and narrowed technical gap, the automotive industry is now likely to go through a similar development process.

Hardware design and performance reached a peak due to the limitation of production process or physical properties of materials, and subjective factors such as consumers' stronger sensitivity and demand for an innovative experience, are driving the automotive industry to seek software technology-based transformation and development, attaching more software value to hardware technology.

The "software-defined vehicles" concept became increasingly prevalent in the automotive industry. Tesla is the quintessential leader of this trend, whose consumer-oriented OTA software services, autopilot packages, differentiated software marketing, as well as the agile undercarriage hardware & software development architecture and central computing platform are the focus of industry research and discussion. Looking at the innovation case of Tesla, the two most remarkable impacts of software-defined vehicles are: first, network functions decoupled from proprietary hardware appliances, enabling parallel physical and digital development of vehicles while software deciding differentiation; second, software becomes commercialized—the new business model represented by Tesla, such as monthly software updates for performance and function improvement, and software subscription like SaaS, maximizes the life cycle and value cycle of vehicles .

Some forward-looking OEMs have begun strategic transformation by enhancing their software capability, while some remain cautious about the software-oriented transformation due to various considerations, such as capital investment and difficulties with internal transformation. This report analyzes the origin of "software-defined vehicles", the driving forces, industry changes, transformation, and new opportunities for the industry, and provides several feasible response models and transformation strategies based on the location and capabilities of the stakeholders on the industry chain, aiming to help OEMs, parts manufacturers, and emerging software companies to comprehend the nature and development process of software-defined vehicles, make rational responses and on-demand layout, and focus on the high profit links of the industry chain transformation in advance.

*"Software-Deﬁned Vehicles—A Forthcoming Industrial Evolution" is independently published by Deloitte, and is not authorized, sponsored or oﬃcially endorsed by Tesla.

1 Software-Defined Vehicles—A Forthcoming Industrial Evolution | 1. Understanding "Software-Defined Vehicles"

1. Understanding "Software-Defined Vehicles"

In 2018 "Software-defined vehicles" Consequently, the core capabilities of Last, enterprises on the industry became a industry hot topic; in 2019, OEMs will shift from mechanical chain, including OEMs and parts Volkswagen CEO Herbert Diess said hardware to electronic hardware and manufacturers, will strengthen that Volkswagen would become a software; the industry value chain will their software capability and software-driven car company1, marking also change from one-off hardware embark on "software-defined the beginning of software-oriented sales to continuous software and vehicles"-centered internal reform industrial transformation. Since then, service premiums. in product development, many interpretations of software- organizational structure, defined vehicles appeared on the First, software and automotive personnel structure, and operation market, including discussions on OTA electronics account for system. In addition, emerging (Over-The-Air) system construction and increasingly more vehicle R&D software companies will capitalize on self-developed operating systems, as costs. The value of in-vehicle the software-hardware synergy to well as detailed analysis of electrical/ software and electronic hardware satisfy the needs of various upstream electronic architecture and basic is expected to exceed that of and downstream enterprises and software platforms. hardware to become the core become the new Tier-1 companies on value of a vehicle. Software cost the automotive industry chain. Based on these interpretations, currently accounts for less than 10% of Deloitte seeks to have a more vehicle BOM (Bill of Material) costs, comprehensive and in-depth which is expected to increase to 50% understanding: "software-defined by 2030—the software includes vehicles" apparently refers to the state application development software, AI that the quantity and value of algorithms, operating systems, as well software (including electronic as the software-hardware integrated hardware2 ) in a vehicle which exceeds controllers, chips and other electronic that of the mechanical hardware; hardware. furthermore, it reflects the gradual transformation of automobiles from Second, software and the highly electromechanical terminals to corresponding improvement in intelligent, expandable mobile performance and functions will electronic terminals that can be determine the differentiation of continuously upgraded. To become future vehicles. Software such intelligent terminals, vehicles are maintenance and upgrading will be the pre-embedded with advanced most economical, convenient, and hardware before standard operating efficient way for future OEMs to procedures (SOP)—the functions and provide differentiated experience and value of the hardware will be gradually improve customer satisfaction. activated and enhanced via the OTA Software iteration will be achieved on systems throughout the life cycle. the basis of hardware redundancy.

1 Source: Volkswagen oﬃcial website, Christian Senger becomes Brand Board of Management Member for new Digital Car & Services function, https://www. volkswagen-newsroom.com/en/press-releases/christian-senger-becomes-brand-board-of-management-member-for-new-digital-car-and-services-function-4664, February 25, 2019. 2 Electronic hardware includes AI chips, microprocessors, domain controllers, etc.

2 Software-Defined Vehicles—A Forthcoming Industrial Evolution | 2. Driving Forces of Software-Defined Vehicles

2. Driving Forces of Software-Defined Vehicles

2.1 Industry development Unlike the Internet industry, the In recent years, smartphone requirements automotive industry features manufacturers have accelerated Software & algorithm— embedded software development, development of in-vehicle infotainment indispensable for the development which means that for each new products, and automobile of connected, autonomous, function, a corresponding ECU manufacturers have increased their shared, and electrified automotive (Electronic Control Unit) will be added R&D investment in intelligence, technologies and new codes will be developed. connectivity, and man-machine With continuous development of Both intelligent connected technology interaction technologies, indicating connected, autonomous, shared, and and autonomous driving require a that consumers' experience and habits electrified automotive technologies, large number of hardware devices on smartphones have expanded to automobiles are transforming at an and involve a huge amount of vehicles, which means that the accelerated pace from mechanical corresponding software development competition focuses with smartphones equipment to highly digitalized and data processing. Therefore, will also be replicated to vehicles. intelligent information-based terminals. in-vehicle software codes increase Many important functions such exponentially. According to preliminary The frequent iteration of smartphones as monitoring and controlling the statistics, for luxury vehicles, the lines allow consumers to experience major battery pack temperature, running of code have exceeded 100 million due improvements in performance and the applications on the center to the high loading rate of Advanced functions through system upgrades, console, human-vehicle interactions, Driver Assistance System (ADAS) and needless of buying the latest phones. and autonomous vehicle detecting L2 autonomous driving. In the next few The OTA is a double-edged model for and classifying the objects around years, the lines of software code are the rapidly iterated consumer can only be achieved by using expected to increase from 100 million electronics. One apparent software and algorithms. Taking to 300 million.3 disadvantage is that consumers are autonomous vehicles as an example: replacing their phones less frequently, an autonomous vehicle is a highly 2.2 Consumer expectations although mobile phone manufacturers software-hardware integrated terminal Consumers expect similar continue to update their products once and the software can be considered behaviors and experience from a year. 4 However, for durable as its "brain" —the "brain" analyzes vehicles as with smartphones. consumer goods that generally have a and utilizes the information collected China has seen the world's highest replacement cycle of at least five years, by various sensors to help the vehicle smartphone penetration rate of nearly the OTA model enables continuous make the best driving decisions. Higher 60%. With ever increasing hardware performance optimization and levels (L3 and above) of autonomous performance, competition of function upgrades throughout the driving will be more complicated smartphones now lies in the software entire life cycle of a vehicle. to require more complex machine ecosystem, innovative application of learning algorithms and deep neural cutting-edge technologies and network models. in-depth exploration of user value.

3 Source: http://news.eeworld.com.cn/qcdz/ic501994.html, July 3, 2020. 4 Source: https://kuaibao.qq.com/s/20190717A0DKYQ00?refer=spider, June 26, 2019.

3 Software-Defined Vehicles—A Forthcoming Industrial Evolution | 2. Driving Forces of Software-Defined Vehicles

2.3 Value chain transfer Apple6 established a strong software software brings, and the new business Accelerated commercialization of ecosystem and sustainable revenue models to provide consumers with hardware enables higher model by virtue of the App Store, the differentiated experience and value. additional value for software Apple operating system, digital Like the smartphone and other products and services. The automotive In Tesla 'case, software services are hardware manufacturing industries, industry is gradually shifting from becoming an increasingly important the automotive industry is undergoing one-off hardware sales to "continuous revenue and profit contributor. In the process of "hardware hardware upgrade, subscription addition to the traditional IoV services commercialization". As the industry service" and other profitable models. (data traffic + in-vehicle content/ signals major technological In recent years, the rise of services), Tesla's software revenue also transformation and complex pre-embedded hardware + FOTA comes from OTA upgrades and automotive electronics emerge, some (Firmware Over-The-Air) model has optional software packages. Tesla traditional mechanical parts are rapidly propelled the OEMs to reconsider the began to roll out paid OTA upgrades in commercialized and become white value of OTA. With pre-embedded 2019—a typical case is that Model 3 labelled products, meaning that performance and advanced hardware, owners can pay $3,000 for the hardware is becoming less the vehicle owners can activate the Acceleration Boost to improve the differentiated and less profitable. As a hidden performance and apply new 0-100km/h time from 4.6s to 4.1s7. The result, the global parts giants with functions through the OTA system more-favored optional autopilot internal combustion engines as their when the algorithms and software package is expected to become the major business segment experienced a mature. The OTA model was valued in main source of income for Tesla' s stock price decline by about 30% over the last round of Internet Of Vehicle software business. Tesla plans to the past three years. However, (IoV) evolution, however, it was limited change the one-time charging model of software and services became more to update of In-Vehicle Infotainment software and offer its Full-Self Driving important on the industrial chain over (IVI) applications and operation (FSD) upgrade via a premium the same period. Autonomous driving interface (i.e., SOTA, Software subscription service. Owners of Tesla full-stack software companies, Over-The-Air), and remote updates did with pre-installed FSD hardware will high-precision map manufacturers, as not bring consumers practical pay only $100/month for the service. well as AI chip and other value-added experience. As FOTA Once the subscription service model is semiconductor enterprises have technology matures, and hardware implemented, all cars with activated created a boom in the capital market. becomes commercialized and FSD are expected to continuously According to preliminary statistics, the standardized, the hardware experience contribute to Tesla's cash flow. autonomous driving enterprises differentiation will quickly disappear. worldwide raised $23.4 billion in 374 OEMs gradually realize that they must investment and financing activities.5 exploit software, the new functions

5 Source: https://kknews.cc/zh-tw/car/e5kk8vz.html 6 "Apple" is the trademark of Apple Inc. registered in the United States and other countries. 7 Source: https://kknews.cc/zh-tw/car/qx4lrvy.html, July 20, 2020. 8 Source: https://auto.sina.com/bg/industry/sinacn/2020-09-21/doc-ihaaeezs2487027.shtml, September 21, 2020. *"Software-Deﬁned Vehicles – A Forthcoming Industrial Evolution" is independently published by Deloitte and is not authorized, sponsored or oﬃcially endorsed by Tesla, Apple or any other companies.

4 otarind hicls—A Forthcoming Industrial Evolution | 2. Driving Forces of Software-Defined Vehicles

Figure 1: Software may become an important contributor to Tesla's revenue

Vehicles sales revenue otar rvnu roduct ictur Speciﬁc functions harging modl Software sales revenue comosition srvics

Autopilot optional •Enhanced autopilot: •One-time charge of $8,000 package (FSD) Auto-park, Navigate on for pre-installation; Autopilot, Smart •The subscription services Hailing, etc. ($100/month) may be available by the end of the year. OTA upgrade •OTA upgrades to constant- •Charge each time for the optional package ly introduce new features speciﬁc update services and improve performance •Online upgrades for powertrain, in-vehicle infotainment, autopilot, electronic, and chassis systems.

2019 2020 2021 2023 2025 Advanced •Advanced IoV connectivi- •Subscription service at IoV Service ty services, including $9.99/month real-time traﬃc, karaoke, streaming, and other functions.

Source: Tesla's Financial Report, Guosen Securities

5 otarind hicls—A Forthcoming Industrial Evolution | 3. Gap between the Present and the Future

3. Gap between the Present and the Future

3.1 The automotive hardware and However, as automotive intelligent and exceed 100 TOPS for L4, which cannot software architecture does not connected technologies continue to be handled by the current computing suit software-deﬁned vehicles. advance, the traditional distributed resources of microcomputers. Another The electrical and electronic EEA with MCU at the core will no longer flaw of the distributed architecture architecture has demonstrated satisfy the development demands of is being unable to share computing constraints in computing power, intelligent vehicles. The pain points power among controllers—the drawbacks in communication include: computing power cannot be shared efficiency, and uncontrolled costs among control modules due to of wiring harness. First, the distributed EEA cannot the one-to-one matching model of The electrification and mechatronics keep up with the increasingly sensors and ECUs; therefore, it is of vehicles have been greatly improved higher computing power. difficult to optimize the distribution since the ECU was first introduced ECU is based on a microcontroller unit of computing power when processing into the automotive industry. The (MCU) and an embedded system— similar functional logic, resulting in a ECU's functions have expanded from the MCU is a microcomputer, and the large amount of wasted computing only controlling engine operation to embedded system is used mainly resources. controlling the chassis, electronic for controlling but not computing. components, as well as in-vehicle Therefore, a single ECU can only infotainment and networking devices, handle computing and control tasks and now each vehicle function is involving a small amount of data, controlled by one or more ECUs. In such as engine control, battery recent years, the number of electronic management, and motor control. In controllers rose significantly with the future, the biggest challenge for the increasing fuel-saving, safety, vehicle development will be the surging comfort, and entertainment demands. higher demand for data processing Currently, more than 100 ECUs and computing speed, either from function on a L2 luxury car. intelligent connectivity or autonomous driving technologies. In particular, the In the early development stage development of autonomous driving of mechatronics, the Electronic & technology will spark off complex Electrical Architecture (EEA) of vehicles logical operation and unstructured adopts a distributed model where data processing scenarios. The the sensors, ECUs and actuators computing of L2 autonomous driving are in one-to-one correspondence, software has already reached 10 TOPS which ensures anti-interference ability (Tera Operations Per Second), and and independence of the system. the computing power is expected to

6 otarind hicls—A Forthcoming Industrial Evolution | 3. Gap between the Present and the Future

Second, another driving force for (laser radar, radar, camera, etc.) needs be redesigned. The complex wiring EEA architecture upgrading is the to complete real-time information harness layout will lead to higher demand for higher communication processing and fusion, which requires mechanical structure cost, thus efficiency and greater bandwidth higher communication bandwidth increasing the overall BOM costs of the capacity. The current EEA is a signal- and transmission rate. The CAN vehicle and affecting the automated based architecture, where signal is bus operates at Mbps, while the production efficiency. transmitted between ECUs through new communication technology, the CAN (Controller Area Network) the Ethernet, allows sensor data bus. The CAN bus is simple, stable, transmission at Gbps. Therefore, no matter the stronger low-cost, anti-interference, and safe, computing power, higher signal and a single-node failure will not transmission efficiency or for vehicle spread to the entire network. However, Last, difficulty in cost control: as weight reduction and cost control, the with more sensors in the vehicle and ECUs and sensors increase in the automotive electrical and electronic higher demand of the smart cockpit vehicle, wiring becomes more hardware architecture needs to for network bandwidth and latency, costly and difficult. For autonomous be changed from the traditional the demand for data transmission driving at L3 and above, more distributed model to the "centralized, will surge and data communication hardware sensors will be deployed in compact and scalable". will need to be completed at a higher the vehicle. In addition to an increasing rate. For example, in an autonomous number of ECUs, the wiring harness driving vehicle, different sensors layout and installation will need to

AAI AI AAI

EEAIE

h A AE A A A IEAI

IEAI otarind hicls—A Forthcoming Industrial Evolution | 3. Gap between the Present and the Future

Figure 2: Automotive EEA Upgrade Roadmap 图：EE架构升级路径图

raditional rsnt Futur

Distributed EEA Centralized (cross-) Centralized Vehicle-cloud domain EEA vehicle EEA computing IVI

Left front Right front Central Power Chassis Body IVI computing Central Form platform computing platform Sensor Actuator Left back Right back

Power ...

Distributed, independently Several maor domains are The central computing Cloud computing functioning ECU formed based on the platform is the top automotive central CAN and LIN bus-based functions of the automotive decision maker the zone computer sensor communication, BCM electronic components, such controllers are formed actuator architecture Character integrated gateway as power, chassis, cockpit, based on the physical istics autopilot, and body domain. locations on the vehicle, Communication network: CAN and act as a gateway to Ethernet distribute data and electric power.

Dedicated sensors, ECUs Centralize the scattered •Simplified harness design Seamless integration of and algorithms, non-syner- ECUs to the domain control- to reduce costs in-vehicle and cloud gic computing power lers for easier OTA upgrades SOA software architecture architecture: vehicle-end leading to redundancy. Higher computing power that supports iteration and computing for in-vehicle Distributed architecture Available for more flexible, expansion of software real-time processing, with Pros & requires a lot of internal higher-rate communication function supplementary cloud Cons communication, resulting in networks computing to provide a significant increase in Higher requirements for non-real-time data wiring harness cost. security mechanism interaction and processing (e.g., IVI) for an intelligent vehicle.

Functional ECUs Domain controller unit one controller unit Central gateway Sensors actuators

Source: Bosch EEA Roadmap; www.shujubang.com

8 otarind hicls—A Forthcoming Industrial Evolution | 3. Gap between the Present and the Future

The earliest parts manufacturers that Thus, the number of ECUs will be At this stage, the EEA layout will be proposed the EEA concept developed greatly reduced, and the functions will actually guided by the wiring harness a roadmap for OEMs to upgrade their be simplified. However, some functions (the physical areas of the vehicle). The electrical and electronic architecture that require low computing power but OEMs, with consideration of (Figure 3). Bosch divides the vehicle high real-time and safety performance modularity, will optimize function EEA into six stages: modularity, will still be controlled by ECUs. classification and integration, and integration, domain centralization, In-vehicle networks such as CAN bus apply the software to the core zone zone centralization, central computing, will be used for intra-domain controllers.10 Finally, the vehicle and vehicle-cloud computing.9 communication, while Ethernet computing resources will be Integration will make each ECU technology will be introduced for concentrated in a few central perform multiple functions, thus communication between domains. computing units to uniformly control reducing the number of single-function the sensors and actuators. controllers. However, the modular Later, "cross-domain fusion" will further closed-type architecture can only meet integrate domain controllers at the If further simplified, the hardware the needs of autonomous driving same function safety and information architecture of the automotive EEA will below L2; it will not satisfy the safety levels. For example, the power, mainly evolve through three stages: computing power, function and safety chassis, and body domains will be integration, domain centralization requirements of higher-level integrated to be the "vehicle control (DCU and MCU), and vehicle autonomous driving. domain" to control the entire vehicle centralization (as shown in the figure). and perform better real-time and At present, OEMs have stepped onto "Domain centralization" can be divided safety functions; the "intelligent cockpit different transformation paths based into two stages: initially, five main domain" will replace the original on their technology and R&D (software domains can be clarified based on which infotainment domain to perform the talent) capabilities, as well as their functional domain the hardware man-machine interaction and T-box relationships with parts manufacturers modules are located, including the integration functions; the "intelligent and cost balance. power domain, chassis domain, body driving domain" will be responsible for domain, infotainment domain, and ADAS perception, planning, and domain. Each domain will be matched decision-making for high-level with a domain controller unit (DCU) with autonomous driving. "Zone (a strong computing power and wider different concept from domain) control range, and usually equipped with centralization" will be a special stage, multi-core processors. and also the earliest prototype of vehicle central computing.

9 Source: https://www.sohu.com/a/426372694_733088, October 21, 2020. 10 Source: https://www.eefocus.com/automobile-electronics/432370, March 25, 2019.

9 otarind hicls—A Forthcoming Industrial Evolution | 3. Gap between the Present and the Future

图Figure：不同主机厂 3: EEA UpgradeEE架构升级路径类型 Paths of Mainstream Automakers

grad aroach haractristics Architctur Eaml

One-step Skip the domain Vehicle centralized EEA Tesla BMW Mercedes-Benz centralization (multi-do- In-vehicle In-vehicle Based on the main fusion) stage to Cloud computer zone central in-vehicle directly to the in-vehicle computing controller calculator computing central computing Self-developed OS Self-developed platform platform (the top Vehicle OTA OS Vehicle OTA decision-maker) to upgrade available upgrade available mobilize the resources at all levels for direct 2019 2021 2024 application/services.

Radical DCU architecture: Domain centralized EEA Volkswagen perform the functions New centralized EEA consisting of vehicle control, Central cross-domain ECU of three in-car application intelligent driving and servers (ICAS) infotainment through Self-developed OS the classic ﬁve-domain Core or three large domain domain controllers. ECU

Step-by-step Steady progress by the Distributed EEA Domestic automotive brands Bosch EEA roadmap and some oint ventures Weak online upgrade capability of software communication architecture still adopting the traditional CAN bus

Source: Soochow Securities, Deloitte Research

According to the planning of automobile satisfied with the progressive upgrade application ecology, and OEMs enterprises, we have summarized approach of Tier-1 enterprises—the domination in developing more and classified the EEA upgrade paths controllers are deployed in accordance advanced software in the future, into three types: first, the "one- with the OEMs' optimal scheme, such speciﬁcally: step" approach: it skips the domain as Volkswagen. Third, the "step-by-step centralization stage and move directly approach": the enterprises are relatively First, the low degree of to in-vehicle central computing. Taking conservative and follow the upgrade modularization and platformation Tesla as an example, the Model 3 path of Tier-1 enterprises. of automotive software has adopts the zone controller + Central impeded centralized dispatching Control Module (CCM) model—the CCM Following the rules of the IT industry, and collaboration of software integrates the ADAS and IVI modules, the development of software will also resources. OEMs' ECUs are usually and the remaining domain functions drive automotive EEA upgrade. At provided by different suppliers. In fact, (directed by the left and right vehicle present, both the distributed EEA and many underlying software controllers body control modules) are deployed by the automotive software development are repetitive, and the code is mainly the physical locations, which significantly model (ECUs use an embedded ensuring the normal operation of the reduces the wiring harness cost of the system, i.e., highly coupled hardware controllers, such as transmitting and vehicle. The zone controller can achieve and software, which are mostly receiving the CAN bus signals, centralized computing as well as material delivered to OEMs by suppliers under scheduling tasks, and reading and cost balance of the vehicle. Second, the "black-box" model) are hampering writing Flash data. the "radical approach": OEMs are not the OTA functions, the expansion of

*"Software-Deﬁned Vehicles – A Forthcoming Industrial Evolution" is independently published by Deloitte and is not authorized, sponsored or oﬃcially endorsed by Tesla, BMW, Mercedes-Benz, Volkswagen or any other companies.

10 otarind hicls—A Forthcoming Industrial Evolution | 3. Gap between the Present and the Future

However, the underlying code cannot standard application programming Under the traditional waterfall be copied or migrated due to the interfaces (APIs) to achieve software development model, the development suppliers' different software stratification. It decouples the of automotive software is divided programming languages and interface upper-level application software and into different parts of work based standards, as well as the high degree the underlying software, improving the on different functional modules. The of dependence of the software on the reusability and expandability of development team of each part focuses hardware. Therefore, ECU software software and reducing the complexity on developing one function, and each development is highly repetitive and and risks of product development. For development stage follows the overall the resources are not efficiently example, application developers can progress plan, which runs in a flow-line utilized. focus only on the specific application model like a waterfall. The relatively separate development parts are likely functions, needless of considering the to cause a "silo effect"—there is only Second, software and hardware differences of the underlying controller internal improvement with each part but are highly nested, and OEMs software. A more profound impact no overall system/platform optimization; cannot conduct extensive and triggered by the decoupled automotive moreover, the development time of each in-depth updates or customized hardware and software is the gradual part is varied, which requires a top-down development. The distributed transformation from the current signal- work structure and a comprehensive software architecture is a based software architecture to a development sequence plan to advance signal-oriented architecture, where service-oriented architecture (SOA). the development in an orderly manner. information transfer between The nature of SOA is that the software This work structure, which separates controllers is achieved via signals, but of controllers, regardless of the the business structure from the the entire system is closed and static, hardware platform and operating organizational structure, requires good and is absolutely defined at the system they reside on, is shared, by internal coordination for overall work compilation stage. Therefore, when the providing abstract services, by other synergy. In addition, the development OEMs want to modify or increase the software components. progress of each part being dependent function definition of a controller, the on each other can easily lead to a instruction of which will invoke the 3.2 The traditional waterfall queue effect—the overall development function of another controller, they software development model has progress will be affected if one part have to upgrade all the necessary major limitations. is delayed. These problems emerge controllers, thus greatly prolonging the Based on the above changes in in practical work: every development development cycle and increasing the technology architecture, in the stage with a long development cycle development costs. context of software-defined corresponds to an independent test vehicles, automobile R&D will stage, which will be verified level-to-level, Therefore, automotive software shift from the traditional waterfall thus the development and test cannot development will follow the development to agile development be conducted simultaneously; each model.During the development development rules of the IT industry stage depending largely on the results process of software-defined vehicles, and introduce the middleware and of the previous stage gives rise to a rigid vehicles will gradually evolve into overall process, leading to high costs and virtualization technologies to achieve an intelligent mobile terminal and long development cycle. All these factors software modularization, and demonstrate more features of a are contradictory to the software- hardware abstraction and consumer electronic product, raising defined vehicles that require the OEMs standardization, thereby further new requirements on the control of to shorten the time-to-market, develop breaking the coupling relationship development costs and development products based on consumer needs, between hardware and software to cycle; meanwhile, product iteration will enable constant iteration, and respond improve EEA flexibility and just begin after the vehicle is delivered quickly to market demands. Therefore, expansibility. Middleware is a to the consumer. The traditional the traditional waterfall development technology that packages software at automobile R&D follows the waterfall model will increasingly demonstrate its different levels of hardware development model to have a linear limitations. dependency to achieve software R&D eco-chain, where product R&D modularization, and defines a series of ends with SOP.

11 otarind hicls—A Forthcoming Industrial Evolution | 3. Gap between the Present and the Future

The agile development model that is streamlined, which is conducive to industry will face the complex mostly used in pure software achieving close coordination and challenges of hardware development development environments (e.g., cooperation and minimizing and multi-suppliers environment. software companies), better fits the management costs; moreover, due to Tesla, a pioneer in agile development requirements of software-defined work flexibility, the development team for software-defined vehicles, has vehicles with its unique features. can interact with customers and quickly provided valuable experience for the Under the agile development model, satisfy the users' needs by providing industry. In addition to the separated the development teams are classified minimum viable products (MVP) and hardware and software development based on product features, and are making continuous innovation and cycles, Tesla pre-designs the hardware completely responsible for their iteration. All these characteristics are and software, fully considers the future function expansion needs at respective feature, including all consistent with the requirements of SOP, pre-installs the standardized functions of the feature, and all teams software-defined vehicles. hardware required for future function have certain freedom and expansion, and subsequently activates decision-making power. When different However, it is important to note that, new functions of the hardware product features involve common unlike pure software development through software upgrades or function functions, the teams will establish environments, the automotive development. For example, Autopilot cross-feature collaboration industry is unique and complex in its pre-installs hardware and the built-in own way. Software-defined vehicles communities to conduct cooperative functions will be sparked off through will ultimately be about combining development and achieve overall OTA updates, moreover, it supports full software and automotive hardware, optimization. In addition, the business lifecycle software updates.11 and organizational structures under which means the application of agile the agile development model are development model in the automotive

11 Tesla’s 8.1 software update brings Autopilot 2.0 cars up to speed, March 30, 2017, TechCrunch, https://techcrunch.com/2017/03/29/teslas-8-1-software-update- brings-autopilot-2-0-cars-up-to-speed/

AAI AI

EEAIE

h A AE A IEAI

IEAI otarind hicls—A Forthcoming Industrial Evolution | 3. Gap between the Present and the Future

3.3 Organizational structure In China, SAIC is one of the few OEMs decreasing, and the domain controllers and talent supply are major that made a strategic shift toward or central computing platform is shortcomings of software-oriented software. Earlier this year, SAIC deployed in a layered or automotive transformation. established the software center, the service-oriented architecture, the OEMs' organizational structure will Z-ONE, which will focus on intelligent future automotive electronic software be fundamentally reshaped: from driving system engineering, software development is expected to be divided a function-oriented structure to a architecture, basic software platform in a layered way. platform development structure. and data factory. This new software Some OEMs have begun organizational subsidiary is expected to scale up to Currently, automotive software talent structure adjustment since the 500 employees by the end of the year, are in short supply. Automotive announcement of software-oriented 1,000 by next year, and 2,200 by electronic software is a branch of transformation. For example, 2023.14 A few other domestic OEMs are embedded software, which is a Volkswagen established the new increasing their positions for software relatively closed industry, so the talent software department, Car.Software, talent, although no independent source is limited. Most of the last June (which became independent software subsidiaries have been set embedded software development this July). The department planned to up. For example, more than 90% of the talent have ﬂowed to tech companies, recruit nearly 5,000 software open positions at the R&D center of and they have little knowledge of engineers, and announced an overall GAC Group are software engineers.15 hardware, automotive engineering and investment of EUR7 billion in software automotive software. These factors architecture over the next 3-5 years.12 Establishing an independent subsidiary have caused the shortage of Toyota, another mainstream for software development allows automotive software engineers. In automaker substantially advancing the greater autonomy, independence and addition, enterprises need to modify software-oriented transformation, vitality. Automotive enterprises that their recruitment and talent announced in this July that it would have only set up a software team management models to cope with the establish a new holding company—the within the group need to further adjust areas where technologies such as Woven Planet Holdings, and two their internal organizational structure autonomous driving that are still operating subsidiaries early next year, and optimize the vehicle software evolving and the business model is which will focus on developing development process. For example, a unclear, such as changing from autonomous driving, new vehicle top-down platform software position-centered to talent-centered operating system, high-definition map development organization shall be model; in the meantime, they should and other state-of-the-art software.13 established to enable be ﬂexible in job duty setting, OEMs are actively bringing in cross-department cooperation, and performance management, and interdisciplinary talent specializing in implement the common strategy or incentive mechanism. software, algorithms, IoV, autonomous work towards the common goal. driving, AI engineering, electronic Automotive engineering development engineering, etc., aiming to accelerate is divided by functional modules, such the adjustment of the current as powertrain, chassis & body, and personnel structure and increase the infotainment, and each module is number of software engineers to developed independently. However, as preserve competitiveness during the automotive EEA is developing software-oriented transformation and towards "domain center" and product innovation. "centralization", the number of ECUs is

12 Volkswagen with new software unit, June 18, 2019, VW Website https://www.volkswagen-newsroom.com/en/press-releases/volkswagen-with-new-software- unit-5092 13 Toyota Research Institute - Advanced Development to Form Woven Capital, an $800 Million Global Investment Fund, September 10, 2020, Toyota Press Release, https://global.toyota/en/newsroom/corporate/33751885.html 14 Source: https://auto.gasgoo.com/news/202007/22I70198617C601.shtml, July 22, 2020. 15 Source: https://auto.gasgoo.com/news/202007/22I70198617C601.shtml, July 22, 2020.

13 otarind hicls—A Forthcoming Industrial Evolution | 3. Gap between the Present and the Future

3.4 Obstacles from the supply For example, the international parts A few other automobile manufacturers chain system enterprises are incapable of localizing with greater dedication and R&D Connection between vehicle and their ADAS solutions and slow in strength have chosen to build their parts enterprises changes from the responding to customer needs. software infrastructure from the tower-shaped vertical relationship Moreover, the parts manufacturers do ground up—they adopt independent to an annular flat relationship. not provide flexible solutions development and vertical integration OEMs can quickly develop and deploy —Mobileye adopted the tie-in sale of for core parts and system upgrades, to product functions with the help of algorithms and chips, which did not achieve technological superiority and Tier-1 suppliers. For example, for R&D support OEMs' customized algorithms, differentiation, which requires investment in primary autonomous thus holding up the OEMs' customized enormous investment. Under the driving (i.e., L2 and lower level and differentiated product independent development model, systems), most manufacturers have development and resulting in OEMs are no longer bound by the adopted the solution provided by the insufficient product innovation. supplies' technological and hardware Tier-1 suppliers due to the high costs performance barriers as well as for all the development tasks and the Therefore, reshaping of the current development cycles, as a result, the lack of technological advantages. The relationship between vehicle and parts most advanced processors will operate international parts giants, with years of enterprises is largely actuated by on the vehicles, and hardware iteration engineering experience, productization OEMs' desire to achieve autonomy and will coincide with software iteration. capability, and cost control experience, control over autonomous driving can provide mature products that technology. In particular, automotive The traditional supply chain meet the vehicle standards—no software is becoming increasingly relationship will change fundamentally, matter for the autopilot perception and important at the ADAS stage, and whether the automobile decision-making parts (e.g., automakers are no longer satisfied manufacturers choose to cooperate millimeter-wave radar, monocular and with the black-box supply model, with the core software and electronic binocular cameras, and other sensors/ instead, they want to define hardware enterprises, or to develop systems) or for the control parts requirements, functions and standards and vertically integrate their own (drive-by-wire systems). Therefore, for from the top down and purchase products. The collaboration between driver assistant systems, OEMs software, hardware and systems OEMs and sub-suppliers will be further generally tend to directly purchase the separately, and hope to break down enhanced, breaking the tower-shaped software (chips, algorithms)-hardware the Tier-1 core technological barriers (from Tier-2 to Tier-1 and then to OEM) (sensors) integrated products and and directly cooperate with the core supply model and eventually forming a solutions from the Tier-1 enterprises. underlying parts enterprises. Driven by flat supply network. At present, a large number of new this emerging trend, in recent years, vehicles are equipped with Mobileye's domestic and foreign OEMs have vision chips for the Advanced Driver established partnership with autopilot Assistance System (ADAS), and most full-stack enterprises, AI chip OEMs obtains the vision module manufacturers, as well as laser radar capability for their L2 autonomous and other sensor enterprises, in the driving solutions from the suppliers. form of equity participation or strategic However, the supplier model has cooperation. gradually shown its drawbacks:

14 otarind hicls—A Forthcoming Industrial Evolution | 3. Gap between the Present and the Future

Figure 4: Changes in Supply Chain Relationship under the "Software-Defined Vehicles" Trend 图：不同主机厂EE架构升级路径类型

ir and ir ulirs ir ulirs Es

Application software HMI software enterprises Pure New software Internet giants Fuel vehicles vehicles AV algorithm software Algorithm software enterprises

Operating system Operating system /middleware enterprises System software Hardware abstraction layer

Hardware

Basic abstractionIntegration, layertesting, software and debugging Basic software enterprises

Traditional parts Networking/comm enterprises unication modules

Hardware Electronic hardware AI chip enterprises

Mechanical hardware

Newly developed Changes in roles on business areas the supply chain

Source: TF Securities, Deloitte Research, public information

15 otarind hicls—A Forthcoming Industrial Evolution | 4. What Are the New Opportunities

4. What Are the New Opportunities

4.1 Industrial value chain under the head-up display and other display open and standardized automotive "software-defined vehicles" trend devices), ADAS software (e.g. adaptive software architecture jointly Taking the intelligent connected cruise control, collision warning), developed by global major automobile vehicles as an example, what changes and body control software. Some manufacturers, parts suppliers, have taken place to the value of application software is supported by and hardware, software as well as the industrial chain in the wake of various algorithms such as the vision electronics industries. the software-oriented industrial and image processing algorithms of transformation. From top to bottom, ADAS software, and the AI or deep The tool suite specifically covers the industrial chain can be divided into learning-based core algorithms of testing, design and R&D, such the pure software layer, basic software autonomous driving software for as ECU software testing and layer, tool software, and electronic environment perception, decision verification, simulation testing of hardware stack. From the value chain planning, and control execution. autonomous driving system. The perspective, the application and electronic hardware stack includes algorithm software as well as the The basic software layer mainly includes communication modules, data storage software-intensive electronic hardware the operating system and framework modules, GPS inertial navigation, at both ends of the value chain have software. Operating systems are divided sensors, chips, and controllers. relatively high industrial added value, into hard real-time and soft real-time which is now the focus of OEMs, systems, which undertake the function parts enterprises, and technology of controlling and distributing in-vehicle companies. The middle basic software hardware resources. There are only a underpins the software-oriented few operating system vendors in the transformation of vehicles. As OEMs industry, thus seeing a high degree of seek to enhance their autonomy and industrial concentration. The framework software capability, the basic software software is a middleware concept, will be more important on in the which abstracts the hardware resources industrial chain. including the operating system, sensors, actuators and computing platform The pure software layer includes to provide a unified basic software application and algorithm software. interface for the upper-level application The application software contains the and algorithm software development, man-machine interaction and interface thus avoiding repetitive development design of the infotainment system due to differences in operating systems (replaced later by the smart cockpit or hardware. AUTOSAR16 is the most that integrates the instrument panel, common middleware solution. It is an

16 Source: Autostar oﬃcial website, https://www.autosar.org/

16 otarind hicls—A Forthcoming Industrial Evolution | 4. What Are the New Opportunities

4.2 Prominent position of software At the distributed ECU stage, OEMs real-time capability and stability. In the platforms only focus on in-vehicle information stage of cross-domain integration and Through clarifying the industrial systems, developing customized IVI central computing platform, domain chain, we believe that three types of operating systems based on the above controllers with varied requirements software and the related tool chain underlying operating systems. For for real-time capability, security and suppliers will play a key role in the example, the two Linux-based open performance will be integrated, or software-oriented transformation of source projects, AGL and Genivi, are merged into a central computing unit, the automotive industry. widely adopted by OEMs—the two which needs a single operating system platforms allow OEMs to directly use with competent real-time computing First, operating system software. the software codes (70%) that have capability and reliable performance. In the trend of "software-defined been developed by the consumer vehicles", the operating system will electronics/communications industry, As the operating systems "sink" to the become increasingly important, and so they only need to develop the bottom of the industrial chain, OEMs will be automobile manufacturers' remaining 30% software for in-vehicle have begun to independently develop focus point for business layout in the scenarios and differentiated their operating systems based on open intelligent connected vehicle field as experiences. This model lowers the source systems—they will no long rely well as the new technology companies' software development threshold and on the suppliers, and share core data, strategic priority. difficultly for OEMs. Some automobile instead, they will control their own enterprises with strong R&D capability software stack, and fix problems or Operating systems can be classified have joined hands with technology add new functions via OTA. In 2019, into the Unix family, the Windows companies to customize their own Volkswagen announced its huge family, the Linux family (including operating systems (such as AliOS) investment in the R&D of the VW.OS Android), and the RTOS family based on open-source operating operating system, aiming to create a (including QNX\Vxworks) according to systems. In the meantime, technology unified OS platform that will be the type of kernel. Different operating companies have created some compatible with multiple underlying systems, with varied advantages, are frameworks/middleware for operating systems to enable intelligent used in different in-vehicle developers based on specific interaction between different domain environments. For example, the IVI requirements, such as DuerOS. controllers and different displays. system often uses the Android system Before that, Tesla developed its as it emphasizes consumer experience However, with the progress of real-time operating system (RTOS) and diversified application ecosystem, software-defined vehicles, operating based on Linux, which supports IVI and the instrument panel mainly systems will become more complex system and ADAS17. adopts the QNX due to its requirement and important. The automotive EEA for high security, although both belong has different requirements for to the smart cockpit domain. With operating systems at different higher real-time and functional safety development stages. For example, in requirements, autonomous driving the stage of "domain centralization", systems mainly adopt the RTOS, with the operating systems are required to currently three mainstream systems: emphasize: openness, compatibility RT-Linux, QNX, and VxWorks. and ecology; as well as security,

17 Source: http://ﬁnance.sina.com.cn/stock/hyyj/2020-08-11/doc-iivhuipn7999112.shtml

17 otarind hicls—A Forthcoming Industrial Evolution | 4. What Are the New Opportunities

Second, middleware: the hardware and software in the domain reached vehicle standards include in-between software of application centralization stage, such as the the BlackBerry QNX Hypervisor, Wind and operating system. The so-called one-core multi-screen multi-system River VxWorks, Green Hills INTEGRITY distributed middleware is to offer a trend of smart cockpit domain. One Muitivisor, Mentor Graphics distributed computing and major trend of the smart cockpit Embedded Hypervisor, and communication framework to shield domain is that the central panel will OpenSynergy (acquired by Panasonic the kernel differences of various gradually integrate with the instrument last year). underlying operating systems and panel, head-up display and other provide standard interfaces and display devices, with microcontrollers protocols for the upper-layer of various hardware integrated into a application developers. single chip. The key demand behind multi-screen integration is cost Adaptive Autosar and ROS are typical reduction, as it is the cheapest way to operating system middleware. run multiple operating systems of Currently, The Autosar R&D tool chain diﬀerent security level on a and basic software are dominated by a System-On-a-Chip (SOC). Moreover, few overseas parts enterprises, each screen corresponds to a diﬀerent including Elektrobit (acquired by the operating system, which means that Continental Group18), Vector and the smart cockpit domain needs to Bosch. Recently, domestic enterprises support multiple operating systems have begun to develop their own tool such as the QNX, Android, and Linux. chain, for example, Huawei's intelligent Therefore, a virtualization platform is driving domain controller MDC is needed on top of the physical compatible with the Autosar standard hardware to support the operation of architecture, and also provides a the operating systems. complete R&D tool suite. Thereby, the virtual machine manager Middleware is becoming increasingly concept of avionics devices was important in the intelligent vehicle era. introduced to the automotive It is expected that intelligent vehicles electronics industry, and the will be using various operating systems AUTOSAR Hypervior19 was created for a long time. As EEA evolves rapidly based on Autosar standard, which is and operating systems become like a software layer running between incrementally complex, the middleware the operating system kernel and the connecting the underlying hardware physical hardware layer, providing a layer and the application layer has a virtual hardware platform for the great development space. operating system to make it more irrelative to the hardware and Third, virtual machine hypervisor. applicable to various platforms. The The application of virtualization current mass-produced products technology is largely triggered by the that are installed with a virtual demand of EEA for separation of machine hypervisor and have

18 Source: Elektrobit oﬃcial website. https://www.elektrobit.com/about/ 19 Source: http://news.eeworld.com.cn/qcdz/article_2018030921931.html, March 9, 2018

18 otarind hicls—A Forthcoming Industrial Evolution | 4. What Are the New Opportunities

4.3 From Tier-2 to Tier-0.5 mass production capacity for their In terms of sales strategy, some chip Future vehicles will be a highly competitivenss20. Seeing from their enterprises have maintained a close mechatronic intelligent device. Being recent moves, chip enterprises are relationship with Tier-1 suppliers, while software-oriented does not mean that demonstrating to the upstream some directly cooperate with OEMs. hardware is negligible. On the contrary, customers their full-stack capabilities For example, for the partnership hardware serves as an important such as software-hardware integration, between Geely and Mobileye, carrier for software to perform its ecological environment, and open tool Mobileye, instead of only supplying functions. Particularly, the chain. For instance, there is a subtle semi-finished components to Tier-1 software-intensive electronic hardware change in NVIDIA' s product strategy, suppliers, will for the first time provide and semiconductor hardware will rake with focus on promoting its open, a complete solution stack including in greater added value and more expandable and customizable product hardware, software, drive strategy and profits on the industry chain. features. Last year NVIDIA introduced control; Mobileye also plans to provide the DRIVE AGX Orin, a software-defined subsequent software update The development trends in recent platform for autonomous vehicles services23. Skipping the Tier-1 years have showed that algorithm and and robots; it is a set of open-source suppliers, the core hardware vendors chip enterprises, originally Tier-2 pre-trained AI models and training will gain higher profits and a greater enterprises, have strengthened their codes; Once the autonomous driving say in the industry. software-hardware co-development vehicle developers adopt this capability, and fully integrated ecosystem, they can freely expand and hardware resource, system software customize models via the NVIDIA AI and functional software, which enables tools, thus improving the stability and them to satisfy diversified upstream capabilities of their autonomous and downstream demands of the driving system21. industry chain; thus, they gradually change from Tier-2 sub-suppliers to Domestic chip enterprises have made Tier-1 or even Tier-0.5 suppliers of similar adjustments. Horizon Robotics' OEMs, holding a key position during latest Journey™ 3 highlights high the development of intelligent performance, low power consumption, connected vehicles. expandability, safety and reliability, as well as its in-depth understanding of Taking AI chip industry as an example, algorithms and solutions for various domestic and foreign chip enterprises application scenarios of autonomous including NVIDIA, Mobileye, and driving; besides, Horizon Robotics Horizon Robotics have made promotes its flexible and open product significant adjustment to their product features, making available to positioning, core competence and customers its algorithm samples, AI sales strategy over the past two years. chip tool chain, and the tool suite Chip enterprises have long been required for application development evaluated by the computing power to help the manufacturers roll out (utilization of computing power), power products and applications at a consumption ratio (TOPS/W), costs and faster pace22.

20 Source: https://auto.gasgoo.com/a/70220737.html, October 14, 2020 21 Source: NVIDIA Introduces DRIVE AGX Orin — Advanced, Software-Defined Platform for Autonomous Machines, December 19, 2019, NVIDIA Website, https:// nvidianews. nvidia.com/news/nvidia-introduces-drive-agx-orin-advanced-software-defined-platform-for-autonomous-machines 22 Source: Horizon Robotics' official website. https://horizon.ai/journey3.html 23 Source: https://new.qq.com/omn/20201009/20201009A074XB00.html, October. 9, 2020 *"Software-Defined Vehicles – A Forthcoming Industrial Evolution" is independently published by Deloitte and is not authorized, sponsored or officially endorsed by NVIDIA, Mobileye, Horizon Robotics, or any other companies.

19 otarind hicls—A Forthcoming Industrial Evolution | 4. What Are the New Opportunities

Figure 5: Changes of the Role of AI Chip Enterprises on the Intelligent Vehicle Industrial Chain Chain

图：ir到ir

ir sulirs ir

Vehicle Cloud Model 2 ... Model n ... OEM Original equipment manufacturers OEM model 1 5G IoT Computing

Ecological Services Tier-1 Application Controller software hardware

Platform In-vehicle computing platform In-vehicle computing platform

Tier-1.5 System Functional Integrated software software solutions

Kernel Vehicle Tier-2 Chip IP ... Tier-1 Sensor Controller ... Tn OEM Model 2 ... Model n OS model 1

Source: Deloitte Research

20 otarind hicls—A Forthcoming Industrial Evolution | 5. How Should Different Enterprises Respond

5. How Should Different Enterprises Respond

5.1 OEMs transform based on We believe that OEMs need to decide a than marginal costs, and strives to rational assessment and their proper transformation path based on improve user experience; the upper capabilities their business scale, R&D strength, left quadrant represents a moderate The global automotive industry began cash flow status, and historical transformation path: OEMs build their software-oriented transformation last burdens. Brands with products for own software teams while actively year. Different from the past mass production and luxury brands cooperating with technology and transformations, the cost and will choose different transformation Internet companies—they still need risk-sharing "alliance model" has been paths, so will fuel vehicle and electric Tier-1 suppliers or newly-rising further broken, which is gradually vehicle manufacturers. Deloitte has software enterprises to provide basic replaced by the vertical integration identified four software-oriented software as well as software-hardware model. In addition, the transformation transformation paths for OEMs architecture solutions before they is unprecedentedly deepened, according to the transformation develop adequate software R&D extending even to product R&D measures some enterprises have taken capability; the lower left quadrant process, organizational structure and and the future development trends. As represents the conservative path: supply chain network of OEMs. shown in the figure below, the four OEMs continue the typical outsourcing paths are separated in four quadrants model to engage Tier-1 suppliers for Volkswagen, a typical traditional OEM, based on OEMs' software strength. software development. has invested heavily in building a Specifically, the right upper quadrant software R&D team (a separate R&D represents the most ambitious and team from the group) and developing thorough transformation path, which is its own operating system and other characterized by the construction of underlying software infrastructure, full-stack technological capabilities becoming a leader among OEMs in from software architecture, software terms of transformation breadth and R&D, to software engineering, as well intensity. Tesla, a representative of new as the active promotion of vehicle manufacturers, has vertical integration; the right lower demonstrated a high degree of quadrant represents a relatively independent development and vertical ambitious but not exhaustive path: it integration for application software, mainly makes breakthroughs in one or operating systems, AI chips and two core technologies, focuses on the in-vehicle central computing platform. areas where marginal income is higher

21 otarind hicls—A Forthcoming Industrial Evolution | 5. How Should Different Enterprises Respond

Figure 6: Four Paths for OEMs' Software-oriented Transformation

Strong

OEMs: establish strategic cooperation with software enterprises while expanding internal OEMs: have independent full-stack (software, R&D team. algorithms, chips, etc.) R&D capability Supplier relationship: OEMs promote software Supplier relationship: skip Tier-1 and co- ecology, which will be practically built by Tier-1 develop subsystems with Tier-2 suppliers suppliers

Establish strategic cooperation Full-stack technological layout with software enterprises

Weak OEM Control over R&D Strong

Focus on making breakthroughs

Deeper tie with Tier-1 software capabilities OEMs' comprehensive comprehensive OEMs' in key areas

OEMs: develop in-house R&D capability in one OEMs: still adopt the black-box model, and or more areas with strategic differences and follow the parts manufacturers' architecture outsource all other R&D tasks. upgrade roadmap Supplier relationship: cooperate directly with Supplier relationship: OEMs promote software core technology vendors through partnership ecology, which will be practically built by Tier-1 or investment and achieve maximum suppliers. autonomy and control.

Weak

Source: Deloitte Research

22 otarind hicls—A Forthcoming Industrial Evolution | 5. How Should Different Enterprises Respond

5.2 With "oppression from both eco-network, and work closely with core ends", parts enterprises should subsystem enterprises to shorten the seek self-transformation adaptation cycle. With the development of connected, autonomous, shared, and electrified In addition, like OEMs, parts automotive technologies, international enterprises are faced with the pressure parts giants have forayed into the of organizational structure adjustment. software field, focusing on: 1) building In July this year, the global parts giant more comprehensive and flexible Bosch announced that it had merged software-hardware integration its automotive electronics and capabilities, and gradually shifting from software businesses departments and outsourcing to self-development of established a new cross-domain core software; 2) adjusting personnel computing solutions division; thus, the structure to better adapt to software entire Car Multimedia division and development rhythms and cycle; 3) parts of the Powertrain Solutions, exploring new business models. Chassis Systems Control, and Automotive Electronics divisions that The strengths of parts manufacturers develop software-intensive, are their integration capability and mass cross-domain electronic systems production experience. However, driven would be brought together in the new by the "software-defined vehicles" trend, division.24 Its competitor, Continental, the core competence of parts planned to adjust its global manufacturers will be "openness", organizational structure in 2019, and "compatibility" and "ecology". Taking the began the adjustment earlier this smart cockpit for example, Tier-1 year25; Under the new structure, suppliers have mature solutions for "Automotive Technologies" as well as domain controllers' embedded software the Powertrain and Rubber development and hardware integration. Technologies are separated, and However, Tier-1 enterprises, driven by "Automotive Technologies" was further the industrial trends or new divided into two business areas, the requirements of OEMs, must open up Autonomous Mobility and Safety their domain controller development to (AMS), and the Vehicle Networking and allow OEMs to customize applications Information (VNI); besides, the Holistic and functions, support heterogeneous Engineering and Technologies was multi-core operating systems as well as established to assume responsibility reusable and portable basic software, for central development activities and accommodate more powerful (basic R&D and future-oriented computing hardware. Therefore, Tier-1 technological development) in the enterprises need to build more automotive sector and support the comprehensive and compatible two business areas AMS and VNI. hardware and software integration capabilities, open up their supply chain

24 Bosch pools its software and electronics expertise in one division with 17,000 associates, July 21, 2020, Bosch Website, https://www.bosch-presse.de/pressportal/ de/en/bosch-pools-its-software-and-electronics-expertise-in-one-division-with-17000-associates-216256.html 25 Source: Oﬃcial website of Continental, https://www.continental-corporation.cn/zh-cn/新闻中心 /大陆集团新闻稿 /大陆集团宣布新管理层架构 -166896?_ ga=2.193074351.389119842.1603956091-778085108.1603956091, March 15, 2019. 26 Source: https://www.sohu.com/a/424505398_115873, October 14, 2020.

23 otarind hicls—A Forthcoming Industrial Evolution | 5. How Should Different Enterprises Respond

Finally, software will significantly affect based on the calculation of upstream the revenue structure of parts raw materials costs as that with manufacturers. The one-off hardware hardware sales. In addition, as a sales model that is based on material system integrator, it is difficult for cost accounting will be no longer parts manufacturers to obtain the applicable. The marginal costs of highest added value during the software development will disappear, industrial chain transformation. Tier-1 and charging of software may be enterprises need to satisfy the conducted in varied modes throughout upstream OEMs' customization the product life cycle. First, before requirements, and buy proprietary mass production, automotive software development tools or systems from revenue mainly comes from the downstream third-party software one-off project development enterprises and pay the engineering charge—the fees charged to Tier-1 costs (Mobileye's earlier business enterprises or OEMs during the R&D model). Software charges cannot be process for delivery of special software standardized due to software's design and customized development, customization property, thus, Tier-1 and the software loyalties—loyalties suppliers will gradually lose pricing charged for the use of the software/ power. Currently, Tier-1 enterprises systems/development tools are still exploring a flexible charging independently developed by the mode that will fit the era of software company. After mass "software-defined vehicles". For II production, software is charged in the example, Bosch has expressed that form of "royalty fees", "technical service its software and hardware mayAAI be AI fees", and FOTA-based software sold separately or sold together in upgrade fees. The royalty fees are the the future.26 EEAIE single-vehicle royalties charged based h on the number of functional module A AE A IPs used by the customer and the IEAI shipments. In the future, automotive IEAI software enterprises may change to sustainable charging modes (like that of water and electricity), such as the SaaS model.

For traditional parts enterprises, the profit model pain point caused by the increase of the software quantity/value ratio is that they cannot accurately calculate the single-vehicle software costs, so they are unable to maintain their gross margin at a stable level

24 otarind hicls—A Forthcoming Industrial Evolution | 5. How Should Different Enterprises Respond

5.3 Facing both opportunities and applications such as IVI and map are Cooperation with core hardware challenges, automotive software more viewed as commercial software enterprises to enter the automotive companies shall create value products, so these software supply network: some large software based on flexible positioning development companies have a weak and algorithm companies started to As mentioned hereinbefore, along with bargaining power. Third, software transform towards hardware-software the restructuring of EEA, software development requires huge initial integration last year, and have acquired enterprises will become increasingly capital investment, but whether the underlying hardware enterprises or important on the software value chain, subsequent scale effect will generate begun to develop their own hardware, including the application software, great economic benefit is expecting to become the new Tier-1 basic software, and safety testing and unpredictable. Last, the lack of a suppliers in the autonomous driving/ verification software. However, software architecture standard in the intelligent connected era, however, automotive software development will industry will be another big challenge for small and medium-sized software inevitably encounter objective facing the automotive software enterprises, the most competitive and challenges. First, as automobile suppliers.27 efficient market-entry strategy is to electronics and intelligent networking deeply cooperate with upstream core continue to advance, and the global Anyway, software enterprises will have electronic hardware vendors. Mobileye market has an increasingly higher more opportunities than challenges. worked closely with Tier-1 enterprises requirement for automobile safety, the Taking into consideration their strength in the beginning to enter the OEMs' international safety standard (ISO and positioning as well as the supply network; smart cockpit software 26262) for automobile electronic and upstream demands, emerging enterprises may access the supply electrical functions has been set up. automotive software companies/ chain through close cooperation with The high safety level of "ASIL-D" Internet companies may focus on the processor vendors. (applicable to ADAS/autopilot systems) following aspects to create value: under this standard has imposed strict access requirements on software Full-stack software capabilities: enterprises' safety technologies and under the trend of software-hardware development techniques. Moreover, decoupling, OEMs with strong R&D vehicle manufacturers strictly select capability will seek semi-autonomous suppliers, requiring not only safe R&D. Therefore, emerging software functions and stable, reliable systems, companies may enhance their but also technical support throughout advantages in full-stack software the life cycle of vehicles. Second, the layout and that in open, compatible automotive software industry lacks a tool chain, to accommodate OEMs' clear pricing model. If charging by value demands of independent development (i.e., pricing based on consumers' as well as cost and risk control. willingness-to-pay for the products or Go-to-market efficiency: for OEMs services) as that of the consumer that have begun to select software electronics industry, it will contradict suppliers through bidding, software OEMs' strategies. Currently, OEMs view companies need to emphasize the ADAS/autopilot technology as their productization and commercialization core assets and capability, and are efficiency of their solutions. willing to pay a high premium. However,

27 Deloitte Insights, "Software is transforming the automotive world", June 18, 2020, https://www2.deloitte.com/content/dam/insights/us/articles/22951_software-is- transforming-the-automotive-world/DI_Software-is-transforming-the-automotive-world.pdf

25 Software-Defined Vehicles—A Forthcoming Industrial Evolution | Conclusion

Conclusion

For enterprises in the automotive The "software-defined vehicles" as well as opportunities for the new industrial chain, "software-defined transformation will spread to every automotive industry players, such as vehicles" is no longer an experimental aspect of the automotive industry, chip suppliers, software suppliers and path. Technology, market, consumers, covering OEMs' vehicle software and Internet companies. The and many other factors are hardware architecture design, product "software-defined vehicles" substantially pushing the development process, development transformation will be an inexorable software-oriented transformation of organization framework, personnel trend driving the development of the the automotive industry. Automobiles, development, as well as the supply automotive industry over the next 5-10 a complex hardware-based chain system and business models of years. All enterprises in the industrial engineering product, will be the entire industry. It may even lead to chain should make thorough increasingly marked by software. restructuring in some areas. During evaluation and forward planning, and the transformation process towards find a suitable path to retain the software-defined vehicles, there will be initiative during the new industrial difficulties and challenges for transformation. traditional automobile manufacturers

AAI AI

EEAIE

h A AE A IEAI

IEAI Software-Defined Vehicles—A Forthcoming Industrial Evolution | Contact us

Andy Zhou Ricky Liu Automotive Sector Leader Automotive Consulting Director Deloitte China Deloitte China Tel: +86 21 6141 1028 Tel: +86 10 8512 4897 Email: [email protected] Email: [email protected]

Golden Liu Yi Zhou Automotive Risk Advisory Leader Automotive Tax & Legal Leader Deloitte China Deloitte China Tel: +86 10 8512 5309 Tel: +86 10 8520 7512 Email: [email protected] Email: [email protected]

Tony Wang Wilson Liu Automotive Financial Advisory Leader Automotive Audit & Assurance Leader Deloitte China Deloitte China Tel: +86 25 5791 6239 Tel: +86 21 6141 1906 Email: [email protected] Email: [email protected]

AAI AI

EEAIE

h A AE A IEAI

IEAI Office locations

Beijing Hefei Shenyang 12/F China Life Financial Center Room 1201 Tower A Hua Bang ICC Building Unit 3605-3606, No. 23 Zhenzhi Road No.190 Qian Shan Road Forum 66 Office Tower 1 Chaoyang District Government and Cultural No. 1-1 Qingnian Avenue Beijing 100026, PRC New Development District Shenhe District Tel: +86 10 8520 7788 Hefei 230601, PRC Shenyang 110063, PRC Fax: +86 10 6508 8781 Tel: +86 551 6585 5927 Tel: +86 24 6785 4068 Fax: +86 551 6585 5687 Fax: +86 24 6785 4067 Changsha 20/F Tower 3, HC International Plaza Hong Kong Shenzhen No. 109 Furong Road North 35/F One Pacific Place 9/F China Resources Building Kaifu District 88 Queensway 5001 Shennan Road East Changsha 410008, PRC Hong Kong Shenzhen 518010, PRC Tel: +86 731 8522 8790 Tel: +852 2852 1600 Tel: +86 755 8246 3255 Fax: +86 731 8522 8230 Fax: +852 2541 1911 Fax: +86 755 8246 3186

Chengdu Jinan Suzhou 17/F China Overseas Units 2802-2804, 28/F 24/F Office Tower A, Building 58 International Center Block F China Overseas Plaza Office Suzhou Center No.365 Jiaozi Avenue No. 6636, 2nd Ring South Road 58 Su Xiu Road, Industrial Park Chengdu 610041, PRC Shizhong District Suzhou 215021, PRC Tel: +86 28 6789 8188 Jinan 250000, PRC Tel: +86 512 6289 1238 Fax: +86 28 6317 3500 Tel: +86 531 8973 5800 Fax: +86 512 6762 3338 / 3318 Fax: +86 531 8973 5811 Chongqing Tianjin 43/F World Financial Center Macau 45/F Metropolitan Tower 188 Minzu Road 19/F The Macau Square Apartment H-L 183 Nanjing Road Yuzhong District 43-53A Av. do Infante D. Henrique Heping District Chongqing 400010, PRC Macau Tianjin 300051, PRC Tel: +86 23 8823 1888 Tel: +853 2871 2998 Tel: +86 22 2320 6688 Fax: +86 23 8857 0978 Fax: +853 2871 3033 Fax: +86 22 8312 6099

Dalian Mongolia Wuhan 15/F Senmao Building 15/F, ICC Tower, Jamiyan-Gun Street Unit 1, 49/F 147 Zhongshan Road 1st Khoroo, Sukhbaatar District New World International Trade Tower Dalian 116011, PRC 14240-0025 Ulaanbaatar, Mongolia 568 Jianshe Avenue Tel: +86 411 8371 2888 Tel: +976 7010 0450 Wuhan 430000, PRC Fax: +86 411 8360 3297 Fax: +976 7013 0450 Tel: +86 27 8526 6618 Fax: +86 27 8526 7032 Guangzhou Nanjing 26/F Yuexiu Financial Tower 40/F Nanjing One IFC Xiamen 28 Pearl River East Road 347 Jiangdong Middle Road Unit E, 26/F International Plaza Guangzhou 510623, PRC Jianye District 8 Lujiang Road, Siming District Tel: +86 20 8396 9228 Nanjing 210019, PRC Xiamen 361001, PRC Fax: +86 20 3888 0121 Tel: +86 25 5790 8880 Tel: +86 592 2107 298 Fax: +86 25 8691 8776 Fax: +86 592 2107 259 Hangzhou Room 1206 Ningbo Xi’an East Building, Central Plaza Room 1702 Marriott Center Room 5104A, 51F Block A No.9 Feiyunjiang Road No.168 Heyi Road Greenland Center Shangcheng District Haishu District 9 Jinye Road, High-tech Zone Hangzhou 310008, PRC Ningbo 315000, PRC Xi'an 710065, PRC Tel: +86 571 8972 7688 Tel: +86 574 8768 3928 Tel: +86 29 8114 0201 Fax: +86 571 8779 7915 Fax: +86 574 8707 4131 Fax: +86 29 8114 0205

Harbin Sanya Zhengzhou Room 1618 Floor 16, Lanhaihuating Plaza Unit 5A10, Block 8, Kailin Center Development Zone Mansion (Sanya Huaxia Insurance Center) No.51 Jinshui East Road 368 Changjiang Road No. 279, Xinfeng street Zhengdong New District Nangang District Jiyang District Zhengzhou 450018, PRC Harbin 150090, PRC Sanya 572099, PRC Tel: +86 371 8897 3700 Tel: +86 451 8586 0060 Tel: +86 898 8861 5558 Fax: +86 371 8897 3710 Fax: +86 451 8586 0056 Fax: +86 898 8861 0723

Shanghai 30/F Bund Center 222 Yan An Road East Shanghai 200002, PRC Tel: +86 21 6141 8888 Fax: +86 21 6335 0003 28

About Deloitte Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited ("DTTL"), its global network of member firms, and their related entities (collectively, the "Deloitte organization"). DTTL (also referred to as "Deloitte Global") and each of its member firms and related entities are legally separate and independent entities, which cannot obligate or bind each other in respect of third parties. DTTL and each DTTL member firm and related entity is liable only for its own acts and omissions, and not those of each other. DTTL does not provide services to clients. Please see www.deloitte.com/about to learn more.

Deloitte is a leading global provider of audit and assurance, consulting, financial advisory, risk advisory, tax and related services. Our global network of member firms and related entities in more than 150 countries and territories (collectively, the "Deloitte organization") serves four out of five Fortune Global 500® companies. Learn how Deloitte's approximately 330,000 people make an impact that matters at www.deloitte.com.

Deloitte Asia Pacific Limited is a company limited by guarantee and a member firm of DTTL. Members of Deloitte Asia Pacific Limited and their related entities, each of which are separate and independent legal entities, provide services from more than 100 cities across the region, including Auckland, Bangkok, Beijing, Hanoi, Hong Kong, Jakarta, Kuala Lumpur, Manila, Melbourne, Osaka, Seoul, Shanghai, Singapore, Sydney, Taipei and Tokyo.

The Deloitte brand entered the China market in 1917 with the opening of an office in Shanghai. Today, Deloitte China delivers a comprehensive range of audit & assurance, consulting, financial advisory, risk advisory and tax services to local, multinational and growth enterprise clients in China. Deloitte China has also made—and continues to make—substantial contributions to the development of China's accounting standards, taxation system and professional expertise. Deloitte China is a locally incorporated professional services organization, owned by its partners in China. To learn more about how Deloitte makes an Impact that Matters in China, please connect with our social media platforms at www2. deloitte.com/cn/en/social-media.

This communication contains general information only, and none of Deloitte Touche Tohmatsu Limited ("DTTL"), its global network of member firms or their related entities (collectively, the “Deloitte organization”) is, by means of this communication, rendering professional advice or services. Before making any decision or taking any action that may affect your finances or your business, you should consult a qualified professional adviser.

No representations, warranties or undertakings (express or implied) are given as to the accuracy or completeness of the information in this communication, and none of DTTL, its member firms, related entities, employees or agents shall be liable or responsible for any loss or damage whatsoever arising directly or indirectly in connection with any person relying on this communication. DTTL and each of its member firms, and their related entities, are legally separate and independent entities.