01.

2020
Computer on wheels / Disruption in
automotive electronics and semiconductors
FOREWORD

F
or more than 120 years, the car was a mechanical means of getting from A to B, an isolated
system whose standout feature was the engine. That's not the case anymore.
Today, as with other products, most automotive innovations are electronic or software
based. Cars have become "connected": customers expect to be "always online"; traffic updates
are automatically beamed into their cars; and electric vehicles communicate with charging
infrastructure. At the same time, features such as "parking by app" or "adaptive cruise control"
are becoming increasingly common. In fact, the car is turning into a system within a system of
systems. Soon, nearly all cars will be connected and part of the Internet of Things, requiring secure
connections for updates and upgrades over the life of the car.
But there is a downside to this progress. The complexity of electronics and distributed software
features has reached an unprecedented level that is proving difficult to handle. New hardware and
software technologies have emerged to help untangle this complexity, but they are not without
their own challenges.
For example, the automotive industry is leveraging advances in computing power to consolidate
electronic control units (ECUs) into centralized computing platforms connected by the automotive
ethernet. This greatly reduces complexity on the network level because of lower relative hardware
costs, but greatly increases software complexity within these processors.
In addition, end-to-end software platforms promise to reduce software complexity by allowing
the "plug-and-play" of new functions and lowering hardware requirements. However, not even
basic software is completely independent of the chipset. Processor performance and possible
parallelization of tasks must be considered when designing software platforms and applications,
as well as in the electrical/electronic (E/E) architecture. The E/E architecture and the geometric
architecture of the vehicle affect each other and cannot be designed independently.
These factors suggest a paradigm shift is required to fully address the complexity. Ultimately,
future vehicle E/E architectures, software platforms and applications need to be designed
around next generation processors – just as vehicles have always been designed with powertrain
performance in mind. In short, electronics hardware and application software will become the
major battlefield for differentiation and control of value creation.

This study is the first in a series on "computers on wheels" by Roland Berger's Advanced Technology
Center. Each aims to address a different aspect of the transformational challenge facing the
automotive industry, with the intention to create line-of-sight for our clients and other interested
parties.

Dr. Wolfgang Bernhart Dr. Michael Alexander

Advanced Technology Center Advanced Technology Center

2 | Focus
 PAGE CONTENTS

4 Management summary

6 1 Introduction
A "computer on wheels"

7 2 MADE to measure
The key disruptive trends in automotive and their consequences
for electronics and semiconductors

11 3 Power change
Three big developments shaking up the electronics
and semiconductor supply chain

17 4 Recommendations
Four steps to tackle disruption in the electronics
and semiconductor supply chain
Cover photo metamorworks/Getty Images

Computer on wheels | 3
MANAGEMENT SUMMARY

Computer on wheels / Disruption in automotive

electronics and semiconductors

T
he car of tomorrow – a computer on wheels – will • Traditional Tier 1 E/E suppliers are under pressure
deliver Mobility as a service, drive Autonomously, from all sides. Their established business models
operate in a fully connected and Digitalized are breaking apart and they risk becoming irrelevant.
environment, and be powered by an Electric drivetrain Furthermore, Tier 1s are at risk of getting stuck with
– MADE. These macro trends are resulting in a significant traditional full front-line liability as an ECU and
increase in the role of automotive electronics and the software stack integrator while losing control over
emergence of the software-driven car, dominated by multiple software components. A
electronics.
The result is that players along the automotive value
The consequences will be dramatic with structural chain are repositioning themselves, and profit pools are
changes all along the automotive value chain: shifting. Semiconductor companies emerge as possible
winners, able to leverage their SoC and SiP platforms to
• OEMs are taking more control over the value chain achieve strong growth in automotive applications.
and critical functionality. They are expanding their
capabilities by adding significant resources for Meanwhile, traditional roles in the supply chain are
module integration, software development and even changing as new players move in and OEMs and Tier 1s
semiconductor design. expand or adapt their focus area to remain competitive.
They are finding they have to understand the business
• Semiconductor players, who frequently control the drivers of semiconductor and pure-play software
largest share of the electronics bill of materials (BoM) in companies to identify suitable win-win positions and
software-driven cars, are moving towards functional mitigate threats. Shorter product and innovation
integration of their chips (system-on-a-chip (SoC), lifecycles are further pressuring them to adapt their
system-in-a-package (SiP), etc). Additionally, they organizational governance and engage in active lifecycle
are expanding from hardware into application-level and innovation roadmap management.
automotive software, as shown by the deal between
Intel and Mobileye.

• Players in electronics contract manufacturing services

(EMS, ODM) are expanding into engineering and
integrated circuits (ICs), offering unparalleled scale
advantages compared to other players.

• Software suppliers that were previously only minor

players in the automotive market are now entering
it all along the value chain, be it in middleware or at
application level.

4 | Focus
A: Value chain reaction
The increased role of electronics and software will mean structural changes throughout the automotive
industry, with Tier 1s particularly affected
ELECTRONICS COMPONENT/IC SUB-SYSTEM/MODULE VEHICLE INTEGRATION

Design IC front Package, Module Module ECU, domain Vehicle

(cmp. IC, end (chip) test design integration and vehicle assembly
SoC, SiP) mfg (boards, ECUs) architecture

BIOS Operating Functional integration Domain & Vehicle to

system, Non-diff. | Differentiating vehicle software cloud
middleware architecture

· Move towards functional

integration by SoC, SiP,
SEMICONDUCTOR more complex boards
MANUFACTURERS · Expand from hardware drivers
middleware to application-
level software

· Move up- and down-stream,

EMS AND ODM
expanding into engineering
COMPANIES and IC back end

· Traditional Tier 1 business

AUTOMOTIVE model under pressure
TIER 1 · Need to reposition in electronics
and software integration

· Leverage differentiated soft-

PURE-PLAY ware capabilities, platforms, and
SOFTWARE service-oriented architectures
COMPANIES · Move into software integration

· Change their strategy towards

more control over the value
chain and critical functionality
· Defend their ownership of E/E
OEMS architecture
· Expand their capabilities for
module integration, software
development & integration

Hardware Software
Source: Expert interviews, Roland Berger Computer on wheels | 5
1 / Introduction
A "COMPUTER ON WHEELS"

T
he shift towards the car of tomorrow – a computer trends in automotive electronics and semiconductors,
on wheels – is already happening, driven by several and makes several recommendations. It is part of a
major changes in the automotive sector. First, series by Roland Berger's Advanced Technology Center
electronic architecture is changing. It is becoming more examining the car as a "computer on wheels". Future
centralized in terms of configuration of computing subjects will include software and IP management.
power with a stronger segmentation of hardware and
software, giving rise to automotive software as a product.
This means the role of all players along the value chain
needs to be redefined.
Second, automotive semiconductors now combine ”The shift towards the
more and more fundamental functionalities. This is
hugely increasing complexity and development costs car of tomorrow –
for embedded software, encouraging semiconductor
companies to move downstream and capture value with a computer on wheels –
higher levels of integration.
Next, power electronics are becoming the is already happening,
differentiator in new electric drivetrains. The technology
and manufacturing base, especially for wide-bandgap driven by several
semiconductors technology (e.g. SiC, GaN), are still
maturing, while OEMs and suppliers need to acquire major changes in the
critical capabilities.
Fourth, the electronics bill of materials (BoM), the automotive sector.”
list of raw materials, components and parts needed to
build a product, is expected to more than double for Falk Meissner
a premium semi-autonomous connected car with an Partner
electric drivetrain1 when compared to a non-autonomous
car with an internal combustion engine. Therefore,
sourcing of electronics is becoming a critical capability
when managing the automotive supply chain.
Lastly, large cross-OEM, cross-player consortia
are forming to mitigate the increasing complexity,
development cost, risk and BoM of future cars, especially
for autonomous driving.
The result of these changes is a fundamental shake-
up of the value chain. To address the challenges facing
OEMs and Tier 1 suppliers, this study assesses recent

1
Excludes battery and drive motor

6 | Focus
2 / MADE to measure
THE KEY DISRUPTIVE TRENDS IN AUTOMOTIVE AND THEIR
CONSEQUENCES FOR ELECTRONICS AND SEMICONDUCTORS

T
he car of tomorrow will deliver Mobility as a
service, drive Autonomously, operate in a fully B: MADE to measure electronics
connected and Digitalized environment, and be Breakdown by MADE factor of the electronics BoM
powered by an Electric drivetrain. The so-called MADE of an ICE vehicle in 2019 and a BEV in 2025 [USD]
trends will be the driving force behind the development
of new automotive technologies.
Despite the proliferation of fleet-based, mobility-
as-a-service vehicles that are autonomous, connected + 3,885 2,235 7,030
and most likely electrified, most cars produced in the
foreseeable future will continue to be individually owned.
Therefore, many of the advances driven by the MADE
trends will occur incrementally via this mass market,
and not via Uber or Waymo robotaxis.
This spells disruption for OEMs and Tier 1 suppliers,
who will need to prepare for the coming changes. In
this chapter we take a closer look at what to expect from 725
each of the MADE trends, and how they will affect the
electronics BoM of future cars. For example, our research
shows that the BoM cost of electronics (i.e. packaged 925

electronics modules) for a mid-sized, premium battery

electric vehicle (BEV) is expected to more than double
to USD 70302 in 2025. This is compared to a similar 3,145 0
standard internal combustion engine (ICE) car today,
where the electronics BoM is USD 3,145. B

MOBILITY – LET’S GO!

Fundamentally, Mobility is a paradigmatic shift from
privately owned and individually operated vehicles to
new business models and use cases. It is not a new
vision – electric taxi cabs were introduced in New
York in 1897. What is new is the potential for mass
adoption, automation of vehicle operation, remote fleet
management, and improved range and reliability of
electric drivetrains. Electric vehicles are now shared and
made available through fleets, for example, and cars are
more tightly integrated with public transportation and
2019
ICE M A D E 2025
BEV

2
Excludes battery and electric drive motor Source: Roland Berger Automotive Electronics Component Model

Computer on wheels | 7
the different means to complete the last mile of travel. of required redundancy for the level of autonomy to be
Nevertheless, there is very little dedicated "mobility achieved. Tesla famously relies on cameras and does not
electronics" in cars; rather, mobility is enabled by use LIDAR (probably for cost reasons), but it remains to
technologies that are developed independently of it. be seen if the industry follows or if LIDAR becomes more
Electric vehicles (EVs) are undoubtedly the future of mainstream once the unit cost decreases with volume. C
mobility, and three key factors are driving electric-vehicle- Autonomous cars rely on an AI-driven central
based mobility: a) simplification of recharging and computing unit that receives and analyzes all raw or
maintenance; b) municipalities pushing for reduction pre-processed sensor data and determines the actions
of urban pollution; and c) regulatory requirements for the car must take. These central computers contain
average fleet fuel consumption. The latter push OEMs multiple advanced chips (SoC ICs) and make up the
to sell electric vehicles into high-utilization applications, other half of the AD-driven BoM increase. But while
such as robotaxis, which can absorb the higher cost. huge progress has been made on chip architectures with
In addition, a significant part of the innovation the advance of massive parallel processors as shown in
required for mobility will be driven by policy makers, chapter 2, challenges remain, especially to reduce power
financial institutions and mobility platforms providing consumption.
amenable business and regulatory environments.
Customer interfaces, financing, payment infrastructure, DIGITALIZATION – ALWAYS ON
insurance, fleet operation, and maintenance will be The car of the future is always on and always connected.
equally important. Already, two thirds of new cars sold in the US have
connectivity features. Connectivity puts the car (and the
AUTONOMOUS DRIVING – RISE OF THE ROBOTS OEM that controls its architecture and data flow) right in
The automotive industry is moving from an era of manual the middle of the automotive IoT ecosystem, with features
driving to one of advanced driver assistance systems such as vehicle tracking, fleet management, maintenance
(ADAS) and autonomous driving (AD). This requires scheduling, over-the-air software updates and remote
shifting from human decision making to artificial access. It also powers autonomous driving, enabling data
intelligence (AI), from memory to maps, and from senses collection, algorithm testing, deployment, and interaction
to sensors. The transition relies on a seamless integration with driving infrastructure and other vehicles (V2X).
of different electronic components, from digitizing The automotive industry is moving towards functional
signals from analog sensors to power electronics for integration to enable new features. In turn, this drives
actuators and drive motors. a growing need to reduce complexity and costs, and a
The biggest technological developments in autonomous fundamental change of the automotive E/E architecture.
driving, and therefore its biggest contribution to growth About a quarter of the rise in electronics BoM in future cars
of the electronics BoM, will be in two areas – sensors and will be attributable to digitalization (USD 725).
computing power. About half of the AD-driven increase
in the electronics BoM (USD 925 in total) is attributable ELECTRIFICATION – THE NEW DRIVETRAIN
to sensors, such as cameras, LIDAR, radar and ultrasonic As noted in Mobility, the automotive future is one of
sensors. The actual figure will depend on the system electric vehicles in one form or another, with several
architecture, pre-processing requirements, and the level factors driving this forward. Regulations tend to push

8 | Focus
towards electrification. In addition to national and than half of the cost increase between the ICE and
state-wide emission standards, an increasing number BEV car is driven by powertrain electrification (battery
of municipalities are banning vehicles with internal management, on-board chargers, converters and
combustion engines. And it appears customers are powertrain inverters), a total of USD 2,235.
increasingly demanding EVs due to their specifications,
lower operating cost and environmental concerns. On MADE’S OVERALL EFFECT
the technical side, higher battery capacities and energy Our research indicates that the share of electronics on
densities are allowing longer driving ranges, while the overall vehicle BoM will increase from about 16%
lower cell costs are reducing the price of EVs. OEMs are for the ICE car to 35% for the BEV, which includes 12%
jumping on the trend by launching large ranges with of powertrain electronics. We estimate that the share
electric/hybrid derivatives of existing models, dedicated of semiconductor components within the electronics
EV models and, in several cases (Daimler, Volvo, GM, BoM will rise from ~25% to ~35%. This will be driven
Volkswagen), even announcing they will stop combustion by the increased complexity and higher integration of
engine development. The impact of high-voltage power semiconductors, as well as an increase in more expensive
electronics on the electronics BoM is significant. More power electronics. D

C: Driving up costs
Autonomous cars rely on a system of sensors and data processing to replace human drivers and make
them safe
SENSE UNDERSTAND ACT

Raw data Object parameters 3D map Actions

GPS, inertial
Pre-processing
measurement

Cameras Pre-processing V2V/V2I Driver monitoring

Radar sensors Pre-processing

Vehicle controls
Sensor fusion Action engine (braking, steering,
etc.)
3D LIDAR Pre-processing

Ultrasonic Offline and Visualization/

Pre-processing
sensors cloud maps display for driver

Source: Texas instruments, Roland Berger

Computer on wheels | 9
D: Electrifying costs
Electronics as a share of total BoM in a 2019 ICE vehicle and 2025 BEV; and cost of semiconductors
as a share of electronics BoM in the same two vehicles

SHARE OF ELECTRONICS BOM IN COMPARISON TO SHARE OF SEMICONDUCTOR CONTENT

TOTAL VEHICLE BOM FOR A PREMIUM VEHICLE [%] IN TYPICAL ELECTRONICS BOM [USD]

100% 100% 7,030

84% 65% Mechanical 4,600 Module

3,145
2,291

12% High power 830 High power

electronics semiconductor

23% Low power 1,600 Low power

electronics ~35% semiconductor
16%
854
~25%

2019 2025 2019 2025

ICE BEV ICE BEV

Source: Roland Berger Automotive Electronics Component Model

10 | Focus
3 / Power change
THREE BIG DEVELOPMENTS SHAKING UP THE ELECTRONICS
AND SEMICONDUCTOR SUPPLY CHAIN

I
t's clear from chapter 2 that the MADE trends will with each other. The CID module includes connections for
have a significant impact on the electronics BoM 5G and other data sources. A central gateway provides a
of future vehicles, especially when it comes to communication channel that allows the OEM to securely
electrification. The technical developments behind these exchange data remotely. This is also used as an interface
trends are too numerous to cover in this report. But to to the powertrain and energy storage.
provide a snapshot, in this chapter we look at three key
developments that will have a significant impact on the Domain fusion and the later step-by-step evolution
electronics BoM – new E/E architectures, autonomous towards a more centralized vehicle computer have
driving chips and new powertrain materials – and several implications for OEMs:
describe their implications. • Domain fusion will be subject to further functional
integration to minimize cross-domain communica-
NEW E/E ARCHITECTURES tion. A split into safety-critical and non-safety-critical
Automotive E/E architectures are set to be simplified functions at the ECU or core level could also emerge.
and consolidated. Rather than comprising up to 120 • Domain controller units (DCUs) will evolve into
separate electronic control units (ECUs), they will centralized computing units. This will increase the
become less distributed and more centralized. At first, complexity and cost of the underlying semiconductor
in the next few years, architectures will be arranged content.
into so-called domains, consisting of several different • Fully and near-fully autonomous driving will require
computing groups fused together. In the longer term, centralized holistic decision making to navigate real-
they will be further centralized into zonal clusters of high- time traffic situations. This is already visible in the
performance computers. Eventually, specific functions available autonomous driving concepts.
will be made cloud based for constant connectivity – the • The new architectures shift complexity from hardware
true computer on wheels. E to software. Applications and functional software will
Different OEMs are taking different approaches to be increasingly separated from electronics, computing
domain fusion. In Germany, for example, it appears hardware and operating systems via standardized
that premium OEMs are focusing centralization on interfaces and software such as AUTOSAR.
five domains: power train & chassis; advanced driver • Hardware will also become more standardized (and
assistance systems (ADAS) & safety; infotainment; commoditized), with the result that automotive software
comfort; and connectivity, in addition to the central will become a product (see our "Computer on wheels/
gateway (the central computing platform or central Software enabled vehicles require software enabled
application server). Development priorities include OEMs and suppliers" study to come). Security of in-
enabling new features, ensuring scalability and vehicle, vehicle-to-vehicle and cloud communication is
improving efficiency to reduce costs. therefore a key issue that needs to be addressed by OEMs.
Over in the US, Tesla has taken a different route. It has • New software and electronics engineering capabilities
split its E/E architecture into four safety-oriented domains: will be required by all players in the ecosystem, in
autopilot; central information display (CID); instrument addition to new (agile) management processes and
cluster; and drivetrain and energy storage. The CID and organizational structures that mirror the high speed
instrument cluster are combined and communicate directly of development cycles in software engineering.

Computer on wheels | 11
E: Centralized solution
The increasing integration of E/E architectures will cut costs and enable new features

DISTRIBUTED DOMAIN-CENTRALIZED ZONAL

Established Next generation (~2021-2025) Future (long term)

Distinctive hardware platforms, Single software platform,

limited HW abstraction full hardware abstraction

TECHNOLOGY
· 80-100 ECUs · 4-5 high-performance DCUs · Cluster of high-performance computers
· Multiple CAN, LIN, Ethernet, · Multiple sensor/actuator ECUs · Multiple sensors/actuator ECUs
Flexray, MOST communication links · 1 CAN bus per zone · 1 CAN bus per zone
· 1 Ethernet backbone · 1 Ethernet backbone

CHARACTERISTICS
· Distributed control · Dedicated domains · Virtualized functions executed in high-
· Many nodes · Consolidation of functions in DCUs performance computers
· Inter-communication via central · Routing between DCUs handled by · Zone-dependent sensors/actuator ECUs
gateway advanced gateway (domain-independent, scalable)
· Routing complexity handled by advanced
gateway

Gateway DCU ECU

Source: Elektrobit; Roland Berger

12 | Focus
AUTONOMOUS DRIVING PROCESSORS • From reference designs to applications, software is
The development of centralized computing architectures becoming an integral part of semiconductor offerings.
for autonomous driving has led to big leaps in the OEMs will therefore increasingly collaborate with
functionality of automotive semiconductors. Automotive semiconductor players, which are traditionally Tier 2
control ICs, which were once mono-functional suppliers.
microcontroller units (MCUs), are now powerful, • Advanced manufacturing nodes for AD processors
integrated autonomous driving processors containing (SoCs) are escalating development costs. This means
multiple processing units (central, graphical, tensor Tier 2 semiconductor players must recoup their
processing units: CPUs, GPUs, TPUs) and dedicated neural development costs across many OEMs. As a result,
processing units (NPUs) with AI accelerators. F only a few will survive as suppliers for AD chips. G
• L arge development ecosystems are emerging to
This rapid evolution has several implications for manage the complexity and share the cost and risk of
automotive players: AD chip development. These collaborations typically
• Semiconductor producers such as Nvidia, Intel, Apple, have one or more OEM and a semiconductor player at
ARM and startups like Intuition Robotics, Hailo Tech their center with Tier 1 and software suppliers adding
and China's HiSilicon are becoming innovation drivers additional capabilities.
as they develop new AI-based architectures. • Managing the semiconductor supply chain will be

F: Hard to process
Automotive control chips have evolved from simple microcontroller units to advanced AI engines

16 BIT 32 BIT 64 BIT

AUTOMOTIVE MCU AUTOMOTIVE MCU AI ENGINE

Legacy product – 1990s Next Generation MCU – 2000s AI engine – 2019

(e.g. ST Micro ST10) (e.g. Infineon Aurix) (e.g. Tesla FSD chip)
· Single core · Up to six Tricore CPUs · 12 CPUS + 2GPU + 2NPU
· 180nm manufacturing node1 · 65 and 90nm · 14nm
· 144 pin package2 · 80-516 pins2 · 2,116 pins
· <10 USD3 typical · 10s of USD3 · 100s to 1000s of USD4
· Monolithic – single task applications · Wide application range: e.g. domain · L4/L5 autonomous driving
control, chassis, safety, engine, ADAS

INCREASING RANGE OF APPLICATIONS, COMPLEXITY OF ARCHITECTURE, MANUFACTURING AND COST

1
Feature size of manufacturing node in nm (nanometers) within the semiconductor structure 2
Depending on version 3
Depending on version and volume 4
Estimate

Source: Company information, data sheets, desk research, Roland Berger

Computer on wheels | 13
G: The chips are up
The smaller the manufacturing node (measured in nanometers), the higher the total development cost
of the processor (SoC) [USD m]

600 EMBEDDED SOFTWARE ACCOUNTS FOR

SOME 30-40% OF DEVELOPMENT COST

500

400
Hardware

300

200

100 Software

0
65nm 40nm 28nm 22nm 16nm 10nm 7nm 5nm

Infineon Mobileye Nvidia Nvidia Mobileye

Aurix EyeQ3 Drive 2015 Drive 2017 EyeQ5 2020
32 bit MCU 2014 (28nm) (12nm) (7nm)
(65nm) (for (40nm)
comparison) Tesla
Mobileye FSD 2019
EyeQ4 2017 (14nm)
(28nm) Nvidia
Drive 2016
(16nm)

Source: Darpa CRAFT, IB5, Gartner, company websites

14 | Focus
 critical for OEMs and Tier 1 suppliers. Semiconductor growth means there is a risk that high-power Si and SiC
components cannot be purchased like conventional semiconductor capacity becomes a bottleneck for the
automotive components, they need to be designed into wider adoption of electric vehicles.
the system. OEMs and Tier 1s therefore need to rethink To prevent this, leading SiC players are making
their long-term position in the electronics value chain. significant investments in the technology. Most notably,
the US firm Cree/Wolfspeed recently announced a USD
NEW POWERTRAIN MATERIALS 1 billion investment in production capacity expansion.
Powertrain electronics, which cover the operation of the But these investments come with their own risks. Market
battery and electric motor in EVs, are developing rapidly, growth is highly dependent on the adoption rate of SiC
with several competing technologies available and no technology in the automotive industry. Technology
industry standard. For example, to enable more powerful breakthroughs in Si-based technologies may mean that
electric drivetrains and faster charging at reduced it continues to compete with SiC. The industry may settle
currents, the industry is moving towards higher voltages. on Si-based technology for mainstream applications due
The 800V module in a Porsche Taycan, for instance, can to the cost advantage and the availability of existing Si
be charged to 80% in less than 30 minutes, much faster production capacity. GaN technology is the wild card in the
than the 200V Toyota Prius device. game, but still has to overcome key physical challenges.
To enable these higher voltage and power
requirements, silicon carbide (SiC) and gallium This uncertainty means OEMs, Tier 1s and
nitride (GaN) technologies have emerged as potential semiconductor suppliers face a series of pain points
replacements for those based on silicon alone (Si IGBT: in the adoption of high-voltage power electronics,
Silicon Insulated Gate Bipolar Transistor). Compared and SiC technology in particular:
to silicon, SiC offers higher power densities, higher • SiC semiconductor costs are 10 times higher than Si-
efficiency and smaller component size, albeit at based technology.
significantly higher cost. H • There is insufficient manufacturing capacity. If SiC
However, managing the high power densities, emerges as the favored technology, existing SiC
currents and thermal loads in these electric powertrains capacity cannot meet the growing demand for EVs.
requires not just rapid development in semiconductors. The release of Tesla's Model 3 alone led to shortages.
Module assembly technology, such as copper bonding, As an additional complication, the manufacturing
sintering and liquid cooling of components, must also be assets for Si and SiC are not interchangeable. Therefore
enhanced, in addition to precise current measurement Tier 2 semiconductor players are reluctant to commit
and control. significant investment without volume commitments
Automotive applications will be the growth driver from OEMs.
for the high-power silicon and SiC markets in the • The electronics industry itself has limited capabilities in
future, where they will compete with renewable energy the handling and manufacturing of the materials, and
applications. It is expected that automotive applications Tier 1s lack skills in the design of the power systems.
will represent about 50% of the SiC semiconductor • The power electronics industry is a virtual oligopoly
market in 2024, which is forecast to grow at ~30% CAGR with Infineon, Mitsubishi and Fuji dominating Si-
to about USD 2 billion at that point. However, this rapid based technology, and Cree/Wolfspeed and Robin

Computer on wheels | 15
H: Raw power
Comparison of the three main types of semiconductors, identified by their key raw material,
used in advanced powertrain electronics

SILICON (SI) IGBT1 SILICON CARBIDE (SIC) GALLIUM NITRIDE (GAN)

· Cost-competitive current · Higher switching speeds & power · Comparable switching speeds to SiC,
standard density for overall higher efficiency/ potentially more cost effective than
· Capacity can be repurposed from reduced losses SiC
other semiconductor segments · Better thermal stability and · Lower package size and higher
conductivity electron mobility compared to Si
· Enables fast charging & longer range · Compatible with existing Si
· Reduces package size, weight and cost manufacturing base

· Least heat resistant; most power loss · Significantly higher cost · Not yet technically mature
· Limited switching frequency · Greater handling & manufacturing · Use in defense applications prevents
difficulty; limited expert pool partnerships between int'l suppliers
· Insufficient SiC manufacturing · High cost with no industrialized
capacity manufacturing base

LONG- Likely to prevail in lower-power Likely to become mass-market To be determined – Market still
TERM hybrids – Potential use for low-end, solution once capacity is built up and highly immature
OUTLOOK low-cost EV volumes drive down cost structure

Source: Roland Berger, desk research 1

Silicon Insulated Gate Bipolar Transistor

ruling SiC-based technology. This limits the supplier • The industry has not yet reached a level of stability,
base. China is attempting to develop a domestic supply maturity or technical capability to suggest the lowering
base, but it is yet to produce significant results. High of prices in the next 5-10 years.
quality standards, requirements for reliability and • Tier 2 semiconductor players are increasingly working
longevity, intellectual property restrictions and capital directly with OEMs, thereby threatening Tier 1s
intensity bar most startups from market entry. position in the value chain.
• The importance of SiC- and GaN-based semiconductors • Investment is likely to stagnate until one of the
for military applications has placed political technologies emerges as the clear winner. This means
restrictions on the export and exchange of technology, strategic partnerships are crucial to accelerate the
know-how and IP. This could potentially disrupt development and adoption of high-voltage powertrain
automotive supply chains and limit the consolidation electronics, and to enable investment in capacity
of the industry, as seen with the blocked takeover of expansion.
Wolfspeed by a non-US company.

16 | Focus
4 / Recommendations
FOUR STEPS TO TACKLE DISRUPTION IN THE ELECTRONICS
AND SEMICONDUCTOR SUPPLY CHAIN

W
ith so much disruption expected in the
automotive electronics and semiconductor
supply chain, no player can afford to sit still. OEMs and suppliers will have specific considerations
But the complexity of the challenges facing them makes when assessing their core value proposition. Depending
it difficult to navigate a path to calmer waters. Most on their starting position, examples include:
importantly, which value chain position should they
take to create sustained competitive advantage? Which Is ADAS truly differentiating or will it be too
partnerships and critical capabilities need to be built up? highly regulated?
And what is the best approach to sourcing electronics
hardware and software, and especially semiconductors? How can we defend a Tier 1 position against new
In this chapter we provide some answers to these platform providers like Nvidia or Mobileye?
questions, as well as making a specific recommendation
for Tier 1 suppliers, who have the most to lose if they fail Is the electric powertrain a differentiator and do
to adapt to the new reality. we need to move powertrain electronics in-
house?

1 DEFINE YOUR VALUE CHAIN POSITION Do we need to control the entire software stack
or only the customer-facing applications?
The first step for any stakeholder in the automotive
electronics and semiconductor value chain is to seriously Do we need to design our own semiconductors?
reflect on their core value proposition. They must identify
differentiating technology capabilities, as well as product Do we build in-house electronics manufacturing
and service offerings that will remain relevant in the long capabilities?
term. In addition, they must choose to defend or expand
a competitive position against the increasing capabilities How do we make money from software as a
of incumbents and new entrants, capture sustainable product – or will software remain an enabler for
profit pools, and decide whether to outsource or rely on hardware sales?
partners.
Not every player will arrive at the same conclusions.
For example, Tesla has chosen to go it alone with an
extreme vertical integration approach to its fully self-
driving capabilities, from the overall system development
all the way to a proprietary semiconductor chip (designed
by an automotive OEM!). Even here, Tesla focuses
on the differentiating capabilities in electronics and
semiconductor design, while leaving the manufacturing
to contractors. I

Computer on wheels | 17
I: The "best chip in the world"1
Tesla's in-house-designed AI engine for autonomous driving

FSD CHIP FSD COMPUTER

IC design Chip manufacturing System design System manufacturing

Software

· In-house IC design · Outsourced · In-house design moving to · Outsourced

· In-house AI software · Samsung Austin 2x Tesla FSD (Gen 3)2 · Quanta Computer,
development (e.g. Deep- · 14 nm node Shanghai, China
scale acquisition)
· IP licensing (e.g. ARM)

In-house Outsourced
1
To quote Elon Musk 2
Previously Nvidia Drive in Gen 2 and Mobileye EyeQ3 in Gen 1

Source: Press research, Roland Berger

2 BUILD CAPABILITIES IN ELECTRONICS, When developing in-house capabilities, the starting point
SEMICONDUCTORS AND SOFTWARE is to leverage existing core competencies. Comparing
these competencies against best practices and/or
No single company can develop the expertise to manage technology requirements will determine gaps and areas
all the elements in the electronics and software supply for potential investment. Electronics, semiconductors
chain, or bear their costs. This is especially true when and especially software have an accelerated development
developing complex systems for autonomous driving, lifecycle, as well as their own development processes.
whose hardware and software requirements, as we have This means they require the structure, agility and
seen above, are highly demanding. processes of a high-tech organization.
As a result, the automotive industry is experiencing a When establishing partnerships or joining a
period of large-scale partnerships and consortia between consortium, participants need to identify the right
rivals. These aim to create a complex ecosystem of diverse partners to close specific capability gaps. They must also
skills, reduce risk, build capabilities and share the decide which partnership option, for example license,
development costs of autonomous driving. Currently two joint development, JV or acquisition, best mitigates the
groups are leading the pack: one consortium involving implementation risks, and balances time-to-market
BMW, FCA, Audi and Intel/Mobileye, and another with with cost, quality, supply security, and the required
Toyota, GM and Nvidia. investment.

18 | Focus
 3 TAKE A STRATEGIC APPROACH TO SOURCING

OEMs and suppliers will have specific considerations Stating the importance of strategic sourcing in the
when developing capabilities or establishing automotive industry seems self-evident. But the sourcing
partnerships in automotive electronics and software. of electronics and semiconductors in automotive is
For example: particularly challenging and costly, for several reasons.
First, quality and production standards are significantly
Do we need a central platform development higher than in other market segments. Automotive
team and how do we structure it? What scale is electronics components have to undergo extensive
needed and how do we achieve it? and time-consuming (i.e. expensive) qualification,3
and component manufacturing must adhere to strict
 Do we need separate technology/software quality processes validated by supplier audits.4 Suppliers,
organizations with a separate budget? How especially those with a consumer electronics background,
do we find, attract and retain the right talent? often do not have the right mindset, qualification or
processes to accommodate this.
How do we manage electronics and software Second, automotive volumes are small in comparison
development versus traditional automotive to other consumer and professional electronics market
development and longer automotive lifecycles? segments. Despite a high per-car electronics content,
additional stock-keeping unit (SKU) fragmentation is
How do we engage with semiconductor players driven by relatively low automotive platform volumes,
(like Nvidia or Intel/Mobileye) if they are already complex E/E architectures with many different components
locked into partnerships? Should we join one and a lack of standardization. Unit volumes are also
of the existing consortia, for example, for self- small in comparison to the semiconductor industry. For
driving technology? example, a 5-million-unit production run for a standard
automotive MCU with 30mm2 die area, which would be
Which startups should we partner with and considered high volume in automotive, corresponds to
what is the best partnership model? How do less than 2 hours of equivalent production capacity of a
we leverage these relationships to build internal 300mm wafer at a semiconductor foundry. J
capabilities? Third, the required ruggedness and reliability is
significantly higher for electronics than it is for consumer
applications. Automotive components must be able to
withstand extreme temperatures and humidity and co-
exist without interference.
Lastly, the lifecycle for automotive electronics
components is significantly longer than in other markets
and can span 20 years. This is in stark contrast to the
3
 .g. Early Life Failure Rate (ELFR), Power Temperature Cycling (PLC), Advanced
E
Product Quality Planning (APQP), Production Part Approval Process (PPAP)
consumer electronics industry, where development and
4
E.g. IATF 16949, AEC-Q 100, AEC-Q 200 product lifecycles can be measured in months.

Computer on wheels | 19
J: Volume game
Unit production volumes in typical electronics markets show how semiconductor costs decrease
as volume increases in 2018

Unit production 1,457 Conceptual cost curve:

[million/year] Decrease expected from scaling of
fixed operating costs (buildings,
IMPLICATIONS facilitation) and high development
FOR AUTOMOTIVE costs, plus indirect costs (e.g.
SEMICONDUCTOR process engineers, management,
indirect chemicals). There is no
"universal" curve as it always
· Despite high per-car
electronics content, depends on the specifics of the
additional SKU process and fabrication.
fragmentation driven by:
– Relatively low
automotive platform
volumes
– Complex E/E
architecture with
many SKUs
– Lack of standardization
~2x
volume · Even in high-volume
automotive production
419 2x
runs, unit numbers are
small in comparison to volume

~2x other market segments 10-15%

volume lower cost
218

95

Passenger LCD Computing1 Smart Volume

and TV phones
commercial
vehicles
1
Tablets, laptops, desktops, servers
Source: Statista, Qorvo, Roland Berger

20 | Focus
The solution? OEMs need to get directly involved 2 – Stay at the cutting edge of the technology to avoid
with semiconductor suppliers when sourcing critical competition on scale. Continuously realign the product
semiconductor components. Advanced procurement portfolio to technology trends and make continuous
is required to build a competitive edge and enable significant investments in innovation.
significant cost savings in automotive electronics. This
should take in supplier management, the optimization of 3 – Dramatically increase scale to compete with global
supply channels and distributors, volume bundling, SKU electronics giants. This path will likely require large-scale
reductions, thinking through minimum order quantities, inorganic growth options.
inventory avoidance and push vs. pull manufacturing
strategies. Equally important in optimizing the BoM
are design for cost and value engineering, feature
optimization, functional integration and platform
strategies from the system level all the way to the chip “Traditional Tier 1
level.
automotive electronics
4 TIER 1 SUPPLIERS NEED TO RETHINK suppliers' position
TO STAY COMPETITIVE
is challenged by
Traditional Tier 1 automotive electronics suppliers are
under pressure from all sides. Their position is challenged up- and downstream
by up- and downstream incumbents and the growing
capabilities of new players. A growing market size and incumbents and the
changing OEM sourcing behavior make the sector both
interesting and accessible to global electronics players, growing capabilities
or contract manufacturers with cross-industry scale. And
as the electronics share of the BoM grows, OEMs' cost of new players.”
pressure will only increase and scale will continue to gain
importance as a competitive advantage. Konstantin Shirokinskiy
This leaves traditional Tier 1s with three options, each Partner
similarly challenging:

1 – Master complete systems, compete on system

knowledge and optimize cost at the system level rather
than component or ECU level. This path requires
integrating system knowledge and design capabilities
with competencies across electronics, software and
mechanical components and architectures.

Computer on wheels | 21
Conclusion

With OEMs and both Tier 1 and Tier 2 suppliers redefining their
role in the changing automotive electronics value chain, three
overarching themes emerge.

F
irst,
 they need to clearly define their area of differentiation
as part of a product and technology strategy based on specific
electronics and/or semiconductor core competencies.
S
econd,
 participation in the electronics value chain needs to
be clearly aligned with these core competencies and the
differentiation areas. Additional partnerships or acquisitions
can help close capability gaps or improve business cases.
F
inally,
 all players can benefit from adapting their organization's
skill set, tools, purchasing and development strategy.
Electronics and semiconductor development and sourcing need
to be value driven and strategic rather than transactional.

Such action may be the difference between future success in

automotive electronics and semiconductors, and irrelevance.

22 | Focus
CREDITS AND COPYRIGHT

AUTHORS

FALK MEISSNER
Partner
+1 510 798-7550
falk.meissner@rolandberger.com

KONSTANTIN SHIROKINSKIY
Partner
+1 248 729-5000
konstantin.shirokinskiy@rolandberger.com

MICHAEL ALEXANDER
Partner
+49 89 9230-8244
michael.alexander@rolandberger.com

We would like to thank Wolfgang Bernhart,

Marcus Baum, Brandon Boyle, Issac Chan,
Stephan Keese, Tara Raizada, Chris Cloutier
and Florian Kalt for their significant contributions

We welcome your questions,

comments and suggestions

WWW.ROLANDBERGER.COM

This publication has been prepared for general guidance only. The reader should not act according to any information provided in this publication without receiving
specific professional advice. Roland Berger GmbH shall not be liable for any damages resulting from any use of the information contained in the publication.
© 2020 ROLAND BERGER GMBH. ALL RIGHTS RESERVED.
ROLAND BERGER, founded in 1967, is the only leading global
consultancy of German heritage and European origin. With
2,400 employees working from 35 countries, we have successful
operations in all major international markets. Our 52 offices
are located in the key global business hubs. The consultancy is
an independent partnership owned exclusively by 230 Partners.

PUBLISHER:
ROLAND BERGER GMBH
Sederanger 1
RB_PUB_19_036

80538 Munich
Germany
+49 89 9230-0