Autonomous Vehicles

Eur

opea

n A

utos

January 26, 2015

Max Warburton (Senior Analyst) • [email protected] • +65-6230-4651

Abbas Ali Quettawala, ACA • [email protected] • +44-207-170-0535

Robin Zhu (Analyst) • [email protected] • +852-2918-5733

Yu-Yu Lin • [email protected] • +44-207-170-0543

See Disclosure Appendix of this report for important disclosures and analyst certifications.

Global Autos: Autonomous Vehicles - It's Even More Complicated Than Realised (An Expert Interview)

Please see the Disclosure Appendix for the ratings and price targets of the companies covered in this report.

Highlights

Autonomous Vehicles ("AVs" - sometimes referred to as “self-driving” or “driverless” cars) are developing rapidly and we are getting an increasing number of questions from investors about what they will mean for the auto industry. The excitement around AVs is understandable – 'newcomers' like Google are making bold claims for their AVs, existing OEMs are demonstrating fast-improving prototypes and suppliers are arguing that they can exploit this new opportunity. People are beginning to ask if AVs are going to fundamentally disrupt the conventional auto industry.

While we also find the technology fascinating and worry about its disruptive potential, we think it is worth highlighting the myriad of obstacles to their adoption – and the difficulty investors will face when trying to make money from their development. While plenty has been written on this subject, we're concerned that the obstacles – technical, operational and legal - are not sufficiently well understood. It is a truism that, while the auto industry eventually adopts very advanced technology, it almost always takes longer to do so than outside observers expect. We believe this phenomenon will repeat itself in the case of AVs.

To help explain the issues, this report takes the form of an interview with industry expert Glenn Mercer, former Director of Automotive Services at McKinsey, who has undertaken extensive research into the progress of AVs. In this report, Glenn explains the massive costs of AV technology, the technical obstacles that remain and the safety, reputational, ethical and regulatory risks – illustrating them with some fascinating anecdotes from those involved in developing these products.

Glenn argues that partially automated driving is realistic, incredibly useful, and inevitable (in fact various vehicles offer some of this now) – but with suppliers, not OEMs, enjoying most of the benefits. However, massive legal, regulatory, and technical obstacles may prevent fully automated AVs becoming realistic for at least another decade, or more. Analyst and investors clearly need to understand AVs. But for OEMs –and therefore for sector investors – AV technology looks likely to be yet another complex R&D cost headwind, rather than something that will drive competitive advantage and dispersion.

∑ We need to understand AVs. Auto sector analysts and investors clearly need to get to grips with Autonomous Vehicles – and quickly. With so much development underway and with the technology evolving fast, this is going to be one of the main R&D priorities of the auto industry. Historically, the auto industry's year used to start with the Detroit Show. But this year one could argue that the more important event was the Las Vegas Consumer Electronic Show, the week before, where a number of OEMs – and their senior management teams – showed up to demonstrate their AV technology. The products on show there were far more interesting and potentially transformational than the performance cars and new variants at Detroit.

∑ How quickly can AVs actually make it to market? AVs are certainly interesting, but how quickly will they become capable enough to be put on sale? Some of the early-stage technology is already being incorporated in cars – although it is not supposed to operate without driver supervision. Some owners have already figured out that the technology is more capable than advertised and, if fooled by a well-positioned and correctly weighted drinks bottle, can actually drive the car. Mercedes-Benz management

Eur

opea

n A

utos

January 26, 2015


2

and their lawyers probably had a collective heart attack when they saw this footage of a Romanian man –and his terrified girlfriend – sitting in the back seat of an S-Class as it drove itself: http://www.liveleak.com/view?i=507_1420300201.

∑ Technology is not the biggest problem – the greater issues may be legal and regulatory. The footage of the over-ridden, self-driving Mercedes is well worth watching – not least because it highlights that the technology itself is not the main obstacle here. While there are still technological hurdles to overcome –Google's technology doesn't work in the snow, for instance - the self-driving car is already a reality, at least in specific conditions such as a highway on a dry day. But there are a huge number of other issues –mainly of a legal, regulatory or reputational issue - that need to be overcome before this technology can become genuinely viable. This report digs into those issues. AV technology is fascinating – and will eventually change significantly the type of cars we drive, their capabilities and just maybe even the companies involved in their production. But in our view the widespread adoption of AVs is still many years off – meaning they are unlikely to change the near to medium term investment thesis on Autos.

Investment Conclusion

We cannot identify a convincing OEM play on AVs. While R&D efforts are ramping up fast, we believe AVs will take longer to get to market and will be more limited in their capabilities than some optimistic forecasts have suggested. Most OEMs are working on this technology and while Mercedes, Volvo, Audiand Nissan have been the most vocal about their technology, it is not clear that any have a clear – or sustainable lead. Newcomers such as Google probably don't plan to become automakers and instead are more likely to license their technology to existing OEMs.

In the long term, AVs may change the economics and balance of power in the industry. But we believe that in the timeframe that most investors in the Autos sector care about – which we presume is less than five years – AV technology will simply be cost headwind (and a conversation topic) rather than a driver of changes in competitive advantage or share prices.

Hard to find listed supplier opportunities yet. Can investors play the supplier space to gain access to AVs? Leaving aside the uncertainty about the speed of AV tech adoption, the problem is finding a listed company with true exposure to this area - other than Mobileye.

The big Tier One integrators like Bosch and Conti will play an enormous role, though these companies are so large that it is hard to see even rapidly-growing AV business moving the needle dramatically for them.Large suppliers who should see more benefit are the safety integrators such as Takata and Autoliv and ZF/TRW, who are in hot pursuit of any and every ADAS feature (you don’t have to believe in high-level AVs to expect to see continued strong growth in ADAS).

Specialist suppliers with perhaps higher leverage to AV roll-out of course include a vision/camera suppliers like Mobileye (listed) - while other firms in this space are typically privately held, such as Cognivue, Omnivue and PMD Technologies. Also in this group are the lidar firms, with Velodyne the acknowledged leader, but there are many others, such as IBEO Automotive in Germany and Quanergy. Sensor companies will be in the mix, with those active including Sensata (listed) to smaller firms like Transense in the UK (listed), IEE in Germany and Elektrobit (listed) in Finland.

Our ratings. We rate VW and Peugeot Outperform in European Autos and Brilliance, Changan and SAIC Outperform in Chinese Autos. These ratings have nothing to do with AVs – and that's unlikely to change for the foreseeable future.

http://www.liveleak.com/view?i=507_1420300201

Eur

opea

n A

utos

January 26, 2015


3

Valuation Comps Table

Exhibit 1Valuation Table

Source: Corporate reports, Bloomberg L.P. and Bernstein analysis.

Details

1. About Glenn Mercer – and the basis of his views and analysis of AVs

In the last year, we've received more and more questions about Autonomous Vehicles (AVs), their potential impact on the companies within our coverage and the opportunities for investors. We have sought outside expertise and this report is based on an interview with Glenn Mercer, former Director of Automotive Services at McKinsey and now independent automotive consultant. Glenn has spent a considerable amount of time exploring the opportunities – and issues – around AVs. We hope you find his views as interesting as we did.

Max Warburton, Bernstein: "Glenn, thanks for making the time to speak today and for sharing your work on AVs. Could you start by just giving us a bit of background about yourself and how you've come look at this space – and how you've gone about trying to research Autonomous Vehicles?"

Glenn Mercer: “As an automotive consultant and analyst of course I couldn’t help taking notice of the upsurge of interest in AVs. The press coverage has been beyond breathless, with some pundits claiming that AVs would not only virtually eliminate hundreds of thousands of traffic deaths worldwide, but free up millions of hours of time otherwise wasted in traffic jams, cut vehicular emissions, liberate shut-ins currently without access to transport, and much more.

But from my own long (and often painful) past experience with suppliers and OEMs working to introduce new technology into cars, I know that these developments are never simple and rarely fast. As the great Japanese automotive researcher Takahiro Fujimoto has said, ‘Cars are born with original sin,’ meaning that, despite our best intentions, vehicles kill people. Accordingly the pace of change is slow: a feature upgrade

22/01/2015 Last Close (Local)

EV/Sales EV/IC EV/EBIT EV/EBITDA

P/E Net Cash(mn)

MV(mn)

EV(mn)

Currency

BMW 98.7 56% 1.0x 5.2x 4.2x 10.6x 14,358 63,079 50,604 EURDaimler 77.6 54% 0.8x 6.4x 4.9x 11.4x 18,136 82,928 75,276 EURFiat 11.0 33% 0.5x 8.2x 3.6x 14.8x -11,440 13,447 33,699 EURRenault 67.5 4% 0.0x 1.1x 0.4x 7.4x 2,316 18,385 3,130 EURPSA 12.1 15% 0.5x 4.6x 2.4x 11.5x 1,962 10,969 9,439 EURVW Pref 198.4 42% 0.6x 6.4x 4.0x 8.0x 20,815 98,309 93,638 EUR

SAIC 23.79 47% 1.3x n.a. n.a. 7.7x 77,600 272,001 211,075 RMBChangan 19.88 57% 2.2x n.a. n.a. 8.4x 214 89,333 87,143 RMBBrilliance 14.48 86% 2.5x 6.5x 4.9x 9.6x -277 58,282 56,333 RMBDongfeng 11.80 32% 0.7x n.a. n.a. 5.8x 1,362 81,424 65,097 RMBGreat Wall 43.10 155% 3.1x 10.7x 9.1x 11.6x 6,813 122,261 115,408 RMBGAC 7.21 30% 1.2x n.a. n.a. 7.3x 10,818 48,918 33,681 RMBGeely 3.12 55% 0.7x 4.6x 3.3x 8.7x 4,512 21,992 17,380 RMBTata 566.50 57% 1.4x 5.2x 3.3x 7.6x -48,913 1,713 1,820 INR (bn)Nissan 1035.00 25% 0.4x 4.2x 2.0x 7.7x 1,134 4,679 3,222 JPY (bn)

Bernstein Global Autos: Valuation Summary - 2015E

Eur

opea

n A

utos

January 26, 2015


4

on a smartphone might fail and cause nothing worse than a dropped call – a failed upgrade to a car’s driving systems might cause a fatal crash.

Accordingly, a few years back I started closely monitoring AV developments, beginning with the DARPA challenges. I knew I could not bring coding expertise to the software side, or an electrical engineering background to the sensors arena, etc., but I felt it would be worthwhile to try to develop an integrated perspective, that cut across all the fields tied up in AVs.

Because make no mistake about it: the modern car is not a stand-alone product, but a key part of a vast system, including OEMs, suppliers, regulators, municipalities, insurers, highway authorities, national and local legal frameworks, and much more. Yet so much of the AV discussion has been just about the narrower “gee-whiz” technical aspects (think George Jetson in his flying car), that AV proponents have tended to either overlook all these other parts of the system or, with perhaps typical Silicon Valley confidence, assume that all that will be taken care of later.

So what I will be discussing with you today is a distillation of what I have been able to learn (after reading countless papers and sitting through hundreds of hours of conference meetings), and more importantly I think, digest and distill into an integrated perspective. I hope this perspective is informed enough to be helpful to your clients (without being so technical as to be incomprehensible), and objective enough to recognize both the vast potential of AVs, and the very real hurdles they have to clear before that potential is realized.”

2. What is meant by Autonomous Vehicles and autonomous driving?

Bernstein: "Glenn, does it make sense to start by actually discussing what we mean by Autonomous Vehicles – this can describe a wide range of technologies and objectives. Perhaps for less specialist readers you could give a recap of what we mean?"

Glenn Mercer: "Good idea – there is no firm definition of this that is entirely stable over time and across borders, though many engineering and regulatory groups are conferring on this. In simple terms, an AV is a vehicle capable of sensing its environment and then navigating (moving) without human input. There are various degrees of such autonomy, so the concept is not binary. AVs are sometimes conflated with “driverless cars,” in which there is no human in control of the car at any point – or even in the car in some cases (like a factory’s AGV). But many industry technicians avoid the term as it ignores the concept of degrees of autonomy, and also conflate two types of highly-autonomous vehicles: one where the human driver need not do anything, and one where there is no human on board at all.”

Levels of AV autonomy

"While there is ongoing debate in the industry about degrees of autonomy, these are the most widely-accepted definitions devised by industry and governmental bodies:

Level 0: No automation. The driver completely controls the vehicle at all times. Today’s default case.

Level 1: Driver assistance. The vehicle may control steering or speed under some conditions, with the driver doing everything else, and continually monitoring the car. Example: active cruise control (ACC).

Level 2: Partial automation. The vehicle may control steering and speed under some conditions, with the driver doing everything else, and continually monitoring the car. Example: automated parallel parking.

Eur

opea

n A

utos

January 26, 2015


5

Level 3: Conditional automation. The driver can fully cede control of all safety-critical functions in certain conditions, but the driver must continually monitor operations and the car must sense when conditions require the driver to retake control, and provide a "sufficiently comfortable transition time" (a very problematic concept, as we will see later) for the driver to do so. Example: traffic jam assistant (for stop-and-go driving).

Level 4: High automation. The vehicle performs all safety-critical functions for the entire trip, with the driver not expected to control the vehicle at any time. Only certain transport conditions are allowed (e.g. highway-only; gated community-only; clear weather only, etc.). Unoccupied car feasible. As opposed to levels 1 through 3, a level 4 car is not yet legal anywhere, except under test conditions.

Level 5: Full automation. Level 4, but with constraints on driving conditions removed. In theory, door to door completely human-unattended operation. I can fall asleep in the back of the car as it takes me to my destination, or I might send it empty to retrieve my child from school. This is obviously what the blue sky scenario is aiming for.”

Exhibit 2Five levels of 'Autonomy' in vehicles

Source: Glenn Mercer LLC and Bernstein analysis.

Other definitions and technologies

Glenn Mercer: "I think I should also mention how ADAS, or 'Advanced Driver Assistance Systems' fit in with this definitional scheme. These are features which assist a driver in his or her tasks. Essentially, higher and higher levels of ADAS shape the implementation of the AV levels. So lane departure warning might come in at level 0, active cruise control at level 1, automatic parking at level 2, traffic jam assistant at level 3, and then a full suite of ADAS systems under AV control can implement levels 4 and 5. It is more than

Level 4

Level 1

Level 2

Level 3

Level 0

Level 5

< No automation. The driver completely controls the vehicle at all times.

< Partial automation. The vehicle may control steering and speed under some conditions, with the driver doing everything else, and continually monitoring the car.

< Conditional automation. The driver can fully cede control of all safety-critical functions in certain conditions, but the driver must continually monitor operations and the car must sense when conditions require the driver to retake control.

< Full automation. Level 4, but with constraints on driving conditions removed. In theory, door to door completely human-unattended operation.

< High automation. The vehicle performs all safety-critical functions for the entire trip, with the driver not expected to control the vehicle at any time. Only certain conditions are allowed (e.g. highway-only; gated community-only; clear weather only, etc.). Not legal.

< Driver assistance. The vehicle may control steering or speed under some conditions, with the driver doing everything else, and continually monitoring the car.

Eur

opea

n A

utos

January 26, 2015


6

likely, therefore, that over time cars will evolve to full AV status through the piling on of more ADAS features, rather than by leaping ahead directly to level 5 in some sort of Kurzweilian ‘singularity.’

What is the 'connected car'?

Furthermore, there is another relevant term of art in play here, the "connected car". There is much confusion around this label, as it can mean something as trivial as a car connected to the internet (to download music), or something more meaningful, such as wireless connection to other cars (to relay speed and trajectory data). In this discussion I will use “connected car” in its narrower definition, as a synonym for a “V2X” car: a car either capable of wireless connection to other cars (V2V, vehicle to vehicle) or to the road infrastructure (e.g. traffic sensors embedded in a highway, V2I, vehicle to infrastructure). (V2X, or vehicle to X, covers both options.) There is an ongoing debate within the AV world as to whether AVs will require V2X to achieve level 4/5, or whether AVs can do this on their own. If there is any consensus, it may be that “Connectivity is not required for AV success, but it will certainly enhance AV operation.”(That is, an AV that relies entirely on maps and sensors to navigate will certainly perform even better if it can also talk to other cars on the road. However, full AV operation using only V2X seems infeasible until virtually all the other cars on the road have it.)”

3. What are the potential benefits of AVs?

Bernstein: Glenn, why are we seeing such a sudden increase in interest in AVs and do you believe they are as close as many people now seem to believe?

Glenn Mercer: "Well the main factor is the march of technology. Technology has matured – by this I mean processing power, sensors, LIDAR (LIght Detection And Ranging), etc. It seems to me that the rising development and use of drones provides some momentum also. There have been a series of DARPA (Defense Advanced Research Projects Agency) AV challenges, where the US military organized several highly-publicized competitions for AV developers that advanced the cause (though intended for military applications, the civilian uses were quite clear). And perhaps most importantly, we have Google aggressively proselytizing, and launching a run of “Google Car” prototypes. But as Google itself has said, we have been predicting AVs since the 1950s, so half a century of failed predictions should keep us humble and realistic.”

Bernstein: Certain followers are rather euphoric about the potential of this technology. Where do you stand on this and what do you think the projected economics might look like?

Glenn Mercer: "Enthusiasts see the benefits as almost infinite, eventually even redefining urban life (fewer traffic collisions, increased roadway capacity, reduced traffic congestion, relief of humans from driving chores, higher speed limits, alleviation of parking scarcity, elimination of redundant passengers (e.g. drivers of delivery vans), reduction in the need for traffic police, reduction in need for vehicle insurance, etc. etc.). However, taking what I consider to be only the categories of benefits that may be viable in the next decade or so, we have:

Benefit 1: Safety

“Virtually all proponents of AVs start here, and they should. Based on the premise that if “90% of all accidents are caused by human error,” (and this does seem to be the case, with tire failure being the biggest factor in the remaining 10%) then in theory removing the human should remove 90% of all accidents. But there are serious challenges to this premise, including the primary one that humans are actually pretty good at driving: in the USA covering some 90 million miles between fatalities (and 65,000 driving hours between

Eur

opea

n A

utos

January 26, 2015


7

an injury-creating accident), as reported by NHTSA (National Highway Traffic Safety Administration). Further, thanks to improvements in roads, drunk-driving laws, and vehicle technologies (e.g. ABS), traffic fatalities are declining virtually everywhere in the developed world -- and so this is a moving target to hit (Germany has gotten down to only 3,000 annual traffic deaths already, and Japan just 4,500). Even one traffic fatality is of course one too many, but one can see the challenge is asking software to better a 1-fatality-in-90-million-miles record.

The safety benefits are indeed enormous: AVs don’t get drowsy at the wheel, don’t drive irresponsibly fast, don’t drink and drive, etc. Broad implementation of AVs will save thousands upon thousands of lives. But human drivers are already setting a very high bar for fully-autonomous (level 4/5) AVs to clear.

Further in this line of thinking, the ADAS implementations that populate levels 1, 2, and 3 of AVs alreadyare getting better every day, and can provide most of the safety improvement that fully autonomous (levels 4 and 5) AVs might offer. Thus one has to consider whether the great expense of going to a level 4/5 AV isthe most cost-effective way to take out the remaining fatalities. Because that expense is indeed great, as I will discuss further: up through level 3 the car only has to defend itself against identified threats (e.g. “you are wandering out of your lane”), whereas at levels 4 and 5 the car has to take on all the other aspects of driving, such as navigation, adherence to traffic regulations, finding a parking spot, turn signal operation, etc., etc. The task becomes exponentially more complex, and thus much more expensive to solve… for what might turn out to be only a modestly incremental savings of lives.

Benefit 2: Convenience

“Many AV proponents next move on to convenience benefits. Based on the premises that (a) we waste a huge amount of time in traffic that is not flowing freely (congestion inconvenience), and that (b) all time spent driving is a deadweight loss (driving inconvenience), to the extent AVs can relieve either or both, they should unlock enormous value. Thus if AVs can route themselves around traffic jams, or safely follow other cars more closely than human drivers could, they might alleviate congestion inconvenience; and regardless of whether there is congestion, if the AV is driving the humans on board can spend their time more productively, perhaps by handling emails (or by playing Angry Birds…).

But there are serious challenges to these premises also. While no one would assert that time spent in traffic jams (especially in the daily commute) is a good thing, note that much of traffic congestion is a price willingly paid by citizens who choose to live at a certain distance from places of work and shopping. That is, if the cost of finding affordable good housing is to endure an extra 30 minutes each day of congestion, many drivers will pay it – and if AV implementation reduces the 30 to 20, they might just then move to even more attractive housing further out. As Ken Laberteaux of Toyota puts it, every improvement in traffic flow we have implemented to date (wider roads, limited access highways, etc.) has just increased urban sprawl: do we think adoption of AVs will have a different effect? Thus we might see commute times change not at all when AVs are widely available.

Driving inconvenience is also not wholly an imposed burden. It is clear that drivers do not consider all driving time a full deadweight loss, that ideally should be completely zeroed out. Academic research by L. S. Redmond and others show that an ideal commute time (in the US) is about 20 minutes (to allow for “alone time,” making the mental transition from work to family, having a coffee, etc.) – which is not far off the average national commute time of 25 minutes. Thus again, we might see consumers assign relatively low value to the reduced driving inconvenience that AVs might bring (see below for a discussion as to why AVs might not actually ameliorate this problem).”

Eur

opea

n A

utos

January 26, 2015


8

Benefit 3: Mobility

“In my view this is the great overlooked benefit, though some experts give it at least lip service. There are (again, using US figures) millions of elderly or disabled persons, whose mobility is currently limited or non-existent, who would welcome the provision of AVs in order to enable them to travel on their own (e.g. to the doctor, to the store). This benefit could be fantastically huge, as the USA population aged 85 and over grows from 6 million today to over 18 million by 2060. Google’s shift in strategic focus from high-speed highway operations to low-speed surface street operations seems to reflect a new focus on this opportunity. Therefore perhaps the greatest social value in AVs lies not in allowing the harried executive to answer three more emails, but in restoring some level of dignity and mobility to the shut-in. And there seems to be no fundamental challenge to the mobility benefit premise. In fact, as this kind of AV service could be provided entirely by LSVs (low-speed vehicles, which are given special regulatory status already, in many nations), total system costs could be much lower (since an AV failure at 150 kph may mean death, but at 50 kph only a bent fender).”

Benefit 4: Commercial

“There is also of course commercial-carrier interest in AVs. This is based on the premise that AVs can offer operators, from delivery services to long-haul transport, various benefits, ranging from “platooning” heavy-duty trucks for lower air drag (and thus better fuel economy), to provisioning of ADAS systems to relieve driver stress and thus permit longer HOS (hours of service). There seems to be no fundamental challenge to the commercial value premise, although one hears grumbles about implementation realities (e.g. how many highway authorities really want convoys of ten “arctics” barreling down the road in a closely-coupled string?).

In a related vein, note also that there has been some talk that at some point a commercial carsharing servicelike Uber (or even a regular taxi company) could switch to level 4/5 AVs, saving thereby the cost of a human driver, and thereby accelerating AV adoption. This is probably quite viable given the low average speed of taxis (average speed of traffic in Manhattan is well under 20 mph), which again mitigates the human cost of any AV failure.”

4. What are the potential costs of AVs?

Bernstein: "Glenn, do you think this technology is affordable – or do you have any ability to quantify whatcosts are going to look like?"

“Well, now we move on to what it will cost to deliver all these benefits (and more). I like to address this by thinking about parts costs, software costs, and the costs of connectivity separately.

The costs of hardware

Parts costs relate to hardware (e.g. sensors, actuators, LIDAR, etc.), and these are mostly falling pretty fast, although the rate varies immensely by AV level and system implementation. (For example, at high speedsAVs can “outdrive” most sensors, so some sensor costs may very well climb before they fall.)

The costs of software development

On the software side - beyond the immense cost of writing the software (tens of millions of lines of code: the average luxury car already has more lines of code than the Large Hadron Collider) -- a huge problem is the cost of test and validation: traditional methods do not work. “Hardware fails by use, but software by

Eur

opea

n A

utos

January 26, 2015


9

design” as the saying goes: that is, software doesn’t wear out, so just running a car in circles for days doesn’t stress it the way it does hardware. Essentially one has to figure out an immense number of conditions which the software might encounter (e.g. “person carrying large box steps into highway at night during rain”) and make sure it can handle each one, which is not as easy as “open and close the sunroof 5,000 times to see if the motor fails.” We have to figure out better software testing schemes.

Millions of hours of software coding and verification needed…

As an illustration, researchers such as Eric Feron of Georgia Tech have asserted that today formally proving and analysing code for correctness takes 0.6 hours per line of code: this is hardly practical for the 50 million line (or more) software systems in AVs. We need new methods: under current testing guidelines, we might have to drive an AV for a decade to wring it out. Fully half the cost of developing the 787 aircraft was software verification – and that was for a vehicle designed to be operated by professional pilots only, and with relatively little risk of collision with other vehicles.

Connectivity costs

Finally, there are connectivity costs: if AVs requires V2X, then various transmitters and receivers (if V2V), or road sensors and transmitters (if V2I) will be needed. These technologies are not necessarily very expensive, but one can imagine the political challenge of getting cash-strapped governments to start embedding sensors in all their roads!

On top of all this are redundancy considerations (e.g. what will happen if the LIDAR and GPS break at the same time?). We will probably need as much redundancy as airplanes, especially as cars do not have pilots who are being paid to stay alert. Redundancy will add expense, as we will likely see vision systems backed up by radar, etc. As one expert has put it, "Sure, LIDAR does a lot, but it can't see if the traffic light turned green."

Thus we might evaluate the marginal cost/benefit analysis of AVs in this way: the incremental costs of moving from level 1/2/3 (ADAS) to level 4/5 (full autonomy) are the highest, while the incremental (safety) benefits may be the lowest.”

5. What are the risks of AVs?

Bernstein: "Glenn, in our previous discussions on this subject you've emphasized your concerns about various risks to which this technology is exposed. Ultimately, you've suggested that the legal risks would probably dissuade big OEMs from rushing this technology to market and that regulators were going to be very, very careful about all of this. What are you current thoughts on these issues?"

Technology Risk

Glenn Mercer: "I think it's important to realize that the technology – at levels 4 and 5, which is where autonomy really kicks in -- is not yet ready for service. Ready for testing, yes; ready for “live” use on limited access highways, yes; but ready to be sold on cars and driven away, no. Let’s just take sensor development issues, as rattled off by engineers (mostly from Tier One suppliers) at a recent AV conference breakout session:

∑ “Sensors generally don’t look far enough ahead yet to provide adequate warning at very high speeds. Bosch personnel have said that today basic sensors can get 70 meters ahead, and high-end sensors 200 meters ahead (at highway speeds). One way around this is to use cooperative sensing

Eur

opea

n A

utos

January 26, 2015


10

(communicating with other cars to get their sensors’ input as it were) and that extends range immensely, but how do we failsafe it? Should an AV ever depend totally on another vehicle?”

∑ “Logging trucks are a problem as their very high ground clearance baffles sensors: the car may try to drive under the trailer.”

∑ “Setting sun ahead of the car screws up many sensor types, rain too.” (See http://goo.gl/mdFQQsfor a video of how this can happen.)

∑ “A sandstorm throws everything out of whack.”

∑ Sensing pedestrians is a big challenge. Of the 34,000 annual USA fatalities, 4,000 are pedestrians. "To date, any system with acceptable performance in pedestrian detection throws off an unacceptable number of false positives."

None of this is to say that the technology will not eventually be ready: these and other problems will eventually be solved. But problems like these both delay roll-out and increase costs, further pushing level 4/5 “Jetson” cars off into the future.”

Disaster Risk

Glenn Mercer: “The costs of things going wrong will of course be huge: as noted, cars kill people. Most dreaded by the industry is the so-called “Hindenburg moment,” in which an AV accident sets the industry back for years. The press has been friendly to AVs so far, possibly because of their ‘wow’ factor, but the press will turn on the industry in a second if it smells blood. One can see the headline now: “FrankenCar Kills Family of Four!” This kind of disaster is related to the next risk category.”

Ethical Risk

Glenn Mercer: “The public is likely not highly tolerant of replacing a great number of human errors (a devil we know) with a smaller number of robot errors: are AV deaths “acceptable” even if they are much fewer in number? Since it is unrealistic to assume no AV is ever in a non-avoidable accident (e.g. simultaneously two cars come at your car, from left and from right), there will be fatalities caused by AVs that are performing perfectly fine. The human driver is usually given a pass in difficult situations like these, as they are under pressure on the spot, where the accident evolves at high speed, so at worst the charge is manslaughter. But for an AV the system programmer, sitting in an office somewhere, and under no immediate pressure, is at fault. Under these conditions, for him the charge may be one of murder. So how good does the AV have to be, relative to human performance, before it can be let off the hook? If AV proponents continue to cite the “90% human error” figure, does that imply AVs will be an order of magnitude better? Thus, one fatality in 900 million miles?

"A Daimler representative provided a useful view on this situation at one industry conference, which I hinted at earlier. If 90% of accidents are due to driver error, then if we take humans out of the loop we slash accidents, by in theory 90%. This is manageable as long as we are at levels 1/2/3, where passive and active safety systems try to “catch” that one fatal error in 90 million miles. But at levels 4/5, where the car become truly autonomous, the statistical challenge inverts. Now, instead of just trying to stop one fatal incident every 90 million, the AV must now also drive the intervening 89,999,999 miles without causing its own accident. ‘So now, during these millions of miles, we have to take on all the other driving tasks, such as obeying all road rules, turning smoothly, avoiding sharp starts and stops, etc. This is very, very hard.’

“One way around this is via a switchable level 4 solution, wherein an OEM makes level 4 a service that one can turn on, but only in the right conditions. If the weather is good, traffic is light, and the highway limited access, we can switch on level 4 for a while, and then switch it off before things get tricky. This allows the AV to run the car only during those miles where its confidence level is high".

Performance risk:

http://goo.gl/mdFQQs

Eur

opea

n A

utos

January 26, 2015


11

Glenn Mercer: “Even if we manage all the other risks, but AV systems just don’t work very well, drivers will be disappointed and enormous sums will have been spent, to little effect. Right now, at least as the pundits are ‘marketing’ AVs, this is almost certain to happen. Media coverage and enthusiasts’ proclamations have led the public to expect level 5 functionality, and soon, whereas only levels 1/2/3 are likely for quite a few years (although forecasts vary widely). Another disappointment could be driven by system parameters, regardless of the AV level. For example, will consumers really want level 4/5, when that means never exceeding the speed limit (because no OEM will ever program a car to break a law)? Or more subtly, will drivers become annoyed when AVs behave as if they were very cautious humans?Everyone reading this note has at some point done a “rolling stop” at a stop sign that (perhaps due to lack of cross traffic) doesn’t really require a complete halt. But an AV doesn’t know that, and more significantly cannot decide to not stop even if it did (thereby breaking the letter of the law), and will come to a full halt. Will this kind of robotic behaviour actually cause accidents, as following drivers are surprised at such conservative behaviour?”

Security risk:

Glenn Mercer: “This risk is getting more and more attention now, as researchers have shown that ‘hacking’ into a car is quite feasible, sometimes through very ingenious entry points such as the tire pressure monitoring system. Researchers Miller and Valasek have even provided a handbook of potential cyber-attack options for a whole range of specific car models, listing both most and least ‘hackable’ models. The nightmare scenario is of course that of some terrorist group figuring out how to (for example) lock the brakes on a thousand AVs simultaneously, potentially through V2X channels. OEMs and suppliers are working hard to deter this, of course, but as computer systems become more complex, their vulnerabilities can multiply.”

Regulatory risk:

Glenn Mercer: “In brief, regulators at the local, state, and national levels may simply not permit as much “progress” as Silicon Valley would like, seeing too much downside (remember our newspaper headline).To date, regulatory bodies in the US in particular have been happy to permit AV testing on the public roads, but regulators have neither the staff on hand nor the expertise (yet) to evaluate AV costs and benefits. Penetration of simpler technologies such as fuel injection, ABS, and airbags reached 100% of developed-world new cars over a matter of decades, not years, in part as regulators cautiously evaluated the benefits and costs. (And infamously, US regulators made errors of judgment as regards the downsides of airbags.) Today’s electronics costs are dropping faster than yesterday’s mechanical costs, certainly, but system complexity is soaring at a rate to offset that fall. Regulators will need to see proven cost/benefit analyses, not just ‘cool features,’ before they grant their approval.”

Liability risk:

Glenn Mercer: “Now we come to the big one, liability. This is of course huge, with multiple moving parts, and with very high potential price tags, which could certainly delay AV roll-out. Examples of the questions regulators, judges, attorneys, insurers, and yes – juries – will be wrestling with include:

1. If AV level 4/5 capability is a feature of the car, does all the liability for its failures shift to the OEM, and out of the hands of the driver? Can the OEM afford this? How can it insure for this? If the average driver pays $1,000 annually per car in insurance, can the OEM afford to take that on? Even if the AV’s insurance premium, thanks to its superior performance, is only $250, will the OEM be responsible for $250 of payments for every car it sells, year in and year out?

2. The single biggest liability issue is managing transfer of control, when the AV decides it needs a human to intervene. How does one set the minimum buffer time to give the human time to orient himself and react (and does this time vary by road condition, driver alertness, and most critically,driver age and skill level)? The analogy some proponents use is from aviation: the autopilot kicks

Eur

opea

n A

utos

January 26, 2015


12

off and returns control to the plane’s pilot, who then corrects the problem. But this is a weak analogy: a plane at 30,000 feet can offer the pilot minutes of time to correct a power failure (e.g.) –a car typically only a few seconds; pilots are trained and paid to pay attention – drivers of level 4/5 cars have put the car into AV mode precisely because they don’t want to pay attention; and pilots are tested for good physical and mental condition – drivers, especially those of advanced age, may not be. (Perhaps the challenge was best put by one ergonomics expert: “We need collaboration, not bail-out-and-hand-over. More like riding a horse than being handed a joystick. The horse generally can follow the trail but we are connected enough to not be startled when all-of-a-sudden it veers off. Unfortunately, we are leading people to dream of being able to completely ignore what the car is doing.")

In short, as long as the driver was in control the bulk of liability passed to him or her: now with the car in control the liability stays with the OEM, and the remedy of handing control back to the driver in a difficult situation seems unlikely to alleviate that.”

Legal risk (beyond accident liability):

“There are questions here that will keep the legal profession busily employed for years. For example: if an AV is speeding, who gets the ticket? More broadly, who is the “operator” if the car is in automatic mode? The OEM? The software supplier? The driver? The person who owns the car? How big an issue is different rule implementation state by state, or nation by nation? How do we handle deliberate joy-ride/misuse of the system? (“I’ll aim the car at the bridge abutment, just to see how late it waits to swerve!”) How do we handle aftermarket AV devices? Some already exist.”

In summary, once again, there may be an imbalance between fully-autonomous (level 4/5) vehicle benefits, and their risks and costs. Again, AV enthusiasts assert that all this will be worked out, but given that firms as strong as Toyota can be laid low by simple floormats jamming the gas pedal, or that OEMs as smart as Ford can see their quality ratings plummet because their telematics suite is a little too confusing, one wonders just how much trouble the much more complex risks of high-level AVs will generate.”

6. What do the various constituencies think of AVs?

Bernstein: "Glenn, there are all sorts of groups with a stake in how this develops – vested interests, newcomers, regulators and the like. What are your thoughts on how they see AVs and what they will do to either drive or retard AV growth?

Transit Authorities

Glenn Mercer: "I'll start with the transit authorities – who are mostly positive on AVs. They are eager to adopt AV buses soon, in part to reduce liability (insurance per city bus driver per year can run $5,000 or more, easily covering the upfront cost of AV equipment). And the routes are mostly predetermined and the speeds slow, so that autonomous operation doesn’t face enormous challenges. There is also interest in level 4 AVs to use for “last mile” applications for disabled passengers, taking them to and from (e.g.) a trolley stop to their home, without the use of expensive paratransit.”

Regulators

Glenn Mercer: "Regulators are obviously going to be critical to this. In theory they are neutral, in actuality both excited and concerned. The DMVs (Department of Motor Vehicles) tend to feel that every increment of AV implementation deskills drivers, which is a worry when systems fail and drivers must take over.

Eur

opea

n A

utos

January 26, 2015


13

Will we therefore need special licenses for AV operation? Take the simple example of backup cameras: they’re a good thing, but do they encourage drivers not to glance back before backing up?

As for Departments of Transportation, the road builders, their questions focus on what kind of roads will beneeded. There is some thinking that we should dedicate lanes for AVs (so they can efficiently platoon at a good speed), but these can generate ruts in the road due to repeated replication of optimal routing, and they can lead to equity issues for those who cannot afford AVs (similar to public resistance to EVs and hybrids using carpool lanes). And then when an AV fails, the usual failure mode is to pull over and stop: but can they do this safely, if the failure mode itself was loss of lane control? On the other hand, DOTs are eager for the data flow that will pour from AVs, that can be used to improve road design, signal timing, etc.

Both of these groups are already arguing, however, that they are neither staffed nor funded adequately to handle the new issues AVs raise. You'll need state and local government support to get AVs on the road. And right now government – at least in the US -- is entirely preoccupied with just keeping the roads we have now maintained (public disgust with deteriorating roads collides with public unwillingness to pay taxes or tolls). Thus AVs are not on any public policy agenda, period – especially if the general public sees them as a toy of the wealthy, which they can use to zoom past all the “regular folk.” If you want to get political attention, you have to show some financial benefit to the town or state, and I have heard none of that. Yes, individual drivers may benefit, but how does this translate into dollars for governments? Or you need some consumer groundswell of support, which we haven’t seen yet. And we may not get it, either, if lurking issues of privacy and government control of AVs in emerge. (Will police be able to remotely stop your car, either for benign reasons such as recovering a driver who has had a heart attack, or for more problematic ones involving arrest on suspicion of a crime?). So right now we have neither the skill nor the will to move, beyond some showy initiatives to approve AV testing.”

Municipalities – no coherent views yet

Glenn Mercer "At the town and city level, the response today is typically “We have no idea yet, the issues are too many and complex.” But there is concern. Widespread adoption of AVs will likely increase urban sprawl (if your commute is more pleasant and productive, why not move further away from work, to a nicer spot in the country), increase VMT (vehicle miles travelled), and probably hurt public transit (“We’re worried that if you can do email in the car, for example, you will move back to back to cars from buses and trains, and clog up the roads”). On a possibly minor point, adoption of AVs could also slash traffic ticket income, which in some small communities is easily half of the annual town budget!”

Congestion – should help, but AVs may initially slow traffic flow as they are more cautious than humans

Glenn Mercer: "While some traffic analysts tell us AVs will improve traffic flow, other researchers feel they may slow traffic, at least in initial implementation, as AVs are more cautious than humans. As one researcher said to me, “If tomorrow all cars were level 5 AVs, traffic would gridlock into a massive jam." This is because, as noted, AVs will scrupulously obey every traffic regulation, and also because, if each car optimizes for its own safety (as it should), for example with 2-second following intervals, then collectively traffic would bog down. To continue with the same researcher (who may not represent the consensus viewpoint): “When we optimize for safety, we cripple flow.” Thus we have to explicitly set the trade-off between flow and safety, realizing perfect safety is achieved only when no one is moving. This problem can only be solved via cooperative traffic management (thus requiring V2X), to make the adjustments for the system as a whole. But will drivers accept this kind of remote control? And is it better centralized (with all the problems thereof) or distributed (ditto)?”

Consumers – will they pay for it?

Glenn Mercer: “Generally consumers are positive about AVs, but not intensely so, and today are more interested in other car features anyway. Surveys also show them nervous about the downsides of AVs. A big problem is that surveys show customers are expecting level 5 (thanks to media coverage), and so will be

Eur

opea

n A

utos

January 26, 2015


14

disappointed when this does not come soon. Also, since many benefits (especially safety) come with levels 1/2/3, will consumers want to pay the significant premium for level 4/5? Various experts have been worrying about this so much that they have dubbed this “the level 4 funding gap.” An analogy might be Concorde: an incredible technological achievement, but not enough people saw the 40% time savings (across the Atlantic) to be worth the 100% ticket premium. It may make sense to adjust our Av marketing to emphasize levels 2, 3, and maybe 4, and move away from level 5 for now.

Glenn Mercer: "I think it is worth noting the J D Power 2014 automotive technology survey findings in this regard. These are, that Americans have:

1. a growing interest in futuristic AVs (level 4/5),

2. but more (and still growing) interest in lower-level ADAS features (traffic jam assist, enhanced ACC, auto parking, emergency brake and steer),

3. excitement about AVs, but also a need for more assurance to become fully confident in them,

4. an interest in AVs -- that is however much lower than for many other things (e.g. smartphone integration), and thus

5. not much willingness to pay a premium (e.g. $1,000 for level 4).

All this doesn’t translate into enormous demand for high-level AVs. That being said, surveys can be weak at prediction, and consumers may alter their views once they see higher-level AVs: who knew they wanted an iPad until they saw one?

Finally, though I have not yet seen surveys regarding “mobility” consumers’ views: would older people welcome a low-speed -- possibly electric AV -- for getting around town? Or would they reject it as stealing their dignity, or signalling to the world that they are not competent? (In the same way that consumers in India tended to reject the Tata Nano, as signalling low status.)”

Academics – thrilled by 'paradigm shifts'

Glenn Mercer: “When it comes to Think Tanks and Silicon Valley Pundits, universally (except for legal professors) – they're obviously enthusiastic. That's not unexpected, since anything that looks like a “revolution” or a “paradigm shift” is thrilling for these groups to pontificate about, seek research grants for, write doctoral theses about, etc. There is a lot of good thinking going on here, but a large share of it is buried under somewhat unmoored theorizing: we certainly can learn from previous (burst) bubbles of enthusiasm for things such as Better Place battery swapping systems, modular car assembly, dealer disintermediation, widespread adoption of build-to-order car manufacture, etc.

Further, there is a lot of irony in some of the thinking in these circles. Libertarian "government can do nothing right" thinkers in the Silicon Valley community keep asking for governments to spend more to enable AVs (e.g. by repainting lane markings on all roads, to be easier for the cars to see) and to do more (e.g. update legislation to permit AVs). But as one industry veteran told me: “It’s hypocritical to attack the government here. One cannot overstate how important in the genesis of the AV field were the original DARPA challenges. Those who habitually question the role of government in commerce, and I know we are sitting here in Silicon Valley, should recognize in this case that they are perhaps wrong. Over time those contests and related AUVSI (Association for Unmanned Vehicle Systems International) programs have generated some 2,500 engineering graduates dedicated to autonomous vehicles."

Media – in love with the concept but will be merciless in the event of accidents

Eur

opea

n A

utos

January 26, 2015


15

Glenn Mercer: “Well, the press is the press and obviously they are interested only in the news value of AVs. And at present, given the possibilities, that coverage is enthusiastic, as the level 5 AV is so futuristic, and the Google Car so odd-looking, etc. But equally obviously the media will “turn on” the industry as soon as there is any kind of dramatic accident or major entrepreneurial implosion. But for now we are on the upward leg of the “hype cycle”.

OEMs – deeply conflicted

Glenn Mercer: “I would say that the OEMs are deeply conflicted. They are happy with the current system of steadily rolling out higher levels of ADAS, very worried about liability issues of level 4/5 AVs, and yet unwilling to appear laggards in their development. They definitely understand that if a few floor mats can cost Toyota a billion dollars or so, and some ignition switches can get GM raked over the coals, that an AV glitch would be devastating. This is especially true given consumers’ proven unwillingness to spend much time understanding advanced systems (see Ford’s problems with MyFord Touch in 2011). Premium OEMs are leading the research charge, as one would expect, though only Tesla seems aggressive enough to be promising some kind of highway level 4 implementation very soon. Other OEMs have announced target dates for various levels of implementation, but the fine print has to be checked (e.g. Volvo’s Drive Me level 4 project rolls out in 2017… but only in Gothenburg… and only for 100 selected drivers.. and only on a few selected roads). So for most OEMs, level 4/5 announcements look more like PR moves – so far.

I have one further comment about OEMs’ outlook in the longer term: the impact of AVs on car ownership levels. Most pundits seem to think that AVs will make car ownership less attractive, as leveraging a fleet of fast-response “Uber AVs” would be more economical than owning your own car, which today is kept parked 95% of the time. Others have argued the reverse, however: if your car has at least level 4 autonomy,you could have it drive you home, drop you off, and go somewhere to park itself. The next morning, you summon it, it picks you up, etc. No need to share it with others, no need to have a parking spot for it at home, etc. This might be very attractive for many people.”

Suppliers – rushing towards the opportunity

Glenn Mercer: “Suppliers are of course interested, but concerned about -- on the upside – how these technologies are monetized, and -- on the downside – keeping out of their traditional automotive sphere interlopers such as MobilEye. The large Tier Ones such as Conti or Bosch will certainly play a major role, for even if new AV technologies emerge that they have not yet mastered, risk-averse OEMs will always want these firms to play the technology integration role, where they have vast amounts of experience. And ADAS firms like Autoliv will of course continue to rise the rising tide of level 1/2/3 systems installations, as the industry continues its evolution from passive to active safety. The “interlopers” such as MobilEye, Elektrobit, or Cognivue certainly see the upside but are wary of the reputation of automotive OEMs for hammering suppliers, and are not particularly excited about working within the traditional tiered supply system. But as we move toward making 100,000,000 cars annually globally, who can resist the volumes?

Insurance Companies – will raise, then lower profitability

These are a special case, with a complex outlook regarding AVs. Right now insurers are undergoing one technology-enabled transformation as UBI (usage-based insurance) rolls out (which will re-allocate customers, and profits, across firms). A second transition underway now and certainly accelerating is the penetration of ADAS, which is reducing the number of collisions. Fewer collisions will probably first improve profits as claims fall faster than premiums, but will eventually reset industry profits at a lower level. And then third are full level 4/5 AVs, which may change the game entirely, in that if liability moves to the OEM, personal driving insurance may evaporate entirely. Why insure the driver if the OEM is the operator of the car?

Eur

opea

n A

utos

January 26, 2015


16

7. How do we monetize AVs?

Bernstein: "Glenn, this is all very interesting and we're glad to get your perspectives on this, but how do my companies – and my clients – make money from this?"

Glenn Mercer: “At present few analysts are discussing how all this will be paid for, possibly just assuming that if the benefits are big enough, somehow the value will flow. The costs of high-level AVs are going to be high and will most likely be paid by the OEM -- whereas the benefits are poorly quantified, and while they may be huge, are for now spread too thin (in that society may benefit more than any car buyer will), to see clearly who will pay for them. But that doesn’t mean there is not value to be mined.

The supplier opportunity

Suppliers of course can see a variety of value-creating opportunities, including:

1. Sensors will be an easy sell for AVs, and of various kinds: optical, infrared, ultrasonic, laser, and beyond…

2. If V2X wireless networks are necessary for AV implementation, there will be transmitter and receiver opportunities (e.g. DSRC (dedicated short-range communication) appliances)…

3. Obviously an enormous amount of software will have to be written, and an enormous amount of testing and validation done…

4. Automatic transmissions (of various sorts) are an absolute must for AVs (I’d apologize to the manual-transmission enthusiasts, but they probably wouldn’t be in the market for an AV anyway)…

5. Navigation systems will be embedded in all AVs, but much of this opportunity has probably already been captured…

6. As AVs required automated controls, makers of sophisticated actuators will benefit. This is a particularly overlooked “nuts and bolts” supplier segment, in part because it does not fit neat categories, like electric motors or sensors, but there are many actuators in a car, from seat adjustors to window lifts, and AVs will require more)

7. The HMI (human machine interface) will be key to customer acceptance, so HMI providers will benefit, probably with no one technology dominant (e.g. large center-stack screens, HUD (head’s up displays), and haptics (such as steering wheels that vibrate when you are drifting out of your lane).

OEMs – probably at a loss about how to profit from AVs

“As for the OEMs, Tesla perhaps excepted, in my view I suspect they are at a loss how to monetize AVs beyond level 3. Up through level 3 we are just adding (ADAS) features, which OEMs are adept at managing, pricing, and bundling. At level 4 and 5, where we enter a qualitatively new realm, it is very unclear how to recover vast development costs. Premium OEMs will of course move first, offering additional ADAS options as they always have, followed by trickle-down (e.g. auto park) to the mass market, probably via a menu of items to make pricing more digestible. But full autonomy may cost so much that it will appear only on the highest-end cars, or will appear first in a dedicated vehicle (much as Toyota chose to introduce hybrid drive in the purpose-built Prius, rather than as a powertrain option, as Honda did with Civic). Certainly a vehicle that is clearly an AV will help boost the status value of the

Eur

opea

n A

utos

January 26, 2015


17

technologies involved, and may even help with liability (as other drivers will see the AV for what it is, and may adjust expectations accordingly, just as they do with heavy-duty trucks or Japanese kei-class cars).

Conversely, an OEM might first roll out level 4 in a small slow city car, maybe a Mitsubishi i-MiEV (perhaps for one’s elderly parents), where the risk profile is lower. Frankly this seems to be what Google is now aiming at. This “low and slow first” strategy frankly could open up the market to recreational vehicle makers such as Polaris and Bombardier, who have relevant expertise. (Note that Polaris has purchased Global Electric Motorcars, an LSV (low-speed vehicle) maker, from Chrysler, in the USA, and Aixam in France.) Certainly US LSV/NEV (neighbourhood electric vehicles) are much more lightly regulated than regular cars.

Further, might an OEM desperate for differentiation break from the pack regarding AVs, as did Toyota with hybrids and Nissan with EVs? Might this be a way for Volvo to regain its long-eroded safety leadership position?

One bright note amidst all this uncertainty: OEMs will likely be able to capture what value there is in high-level AVs, to the extent that value exists. That is, as opposed to the debacles of in-car entertainment, phone services, and even navigation, where value was siphoned off by aftermarket audio systems, personal navigation devices, smartphones, and services like Google Maps and Spotify, high-level AV technology will be so embedded in the car’s basic functionality that it is unlikely we will see the Apples or Verizons of the world “run off” with AV value created. (Though for a rebuttal, see http://www.getcruise.com/.)”

8. Which firms might monetize first?

Bernstein: “But let’s get specific now. Who are the firms, that investors might look at, that will reap value from AV implementation, at whatever rate it occurs?"

Tier One integrators, large suppliers and specialist firms

Glenn Mercer: "As far as suppliers go, of course the big Tier One integrators like Bosch and Conti will play an enormous role, though these companies are so large that it is hard to see even rapidly-growing AV business moving the needle dramatically for them. Large suppliers who should see more benefit are the safety integrators such as Takata and Autoliv and ZF/TRW, who are in hot pursuit of any and every ADAS feature (and as noted, you don’t have to believe in high-level AVs to expect to see continued strong growth in ADAS). Specialist suppliers with perhaps higher leverage to AV roll-out of course include vision/camera suppliers like Mobileye (listed) - while other firms in this space are typically privately held, such as Cognivue, Omnivue and PMD Technologies. Also in this group are the lidar firms, with Velodyne the acknowledged leader, but there are many others, such as IBEO Automotive in Germany and Quanergy. Sensor companies of all kinds will be in the mix, from giants like Sensata (listed) to smaller firms like Transense (listed) in the UK and IEE in Germany and Elektrobit (listed) in Finland are also of interest of course. And we haven’t even touched the app firms like Vinli or Navdy, which don’t actually fit into the definition of AV suppliers, but whose smartphone apps are designed to appeal to the driver who wants a more “intelligent” driving experience as soon as possible. (A related field is that of UBI (usage-based insurance, as pioneered by Progressive), which is only tangentially impacted by AV technology now (ADAS penetration of course reduces accident claims), but over time of course insurers may find high-level AVs putting them out of a job completely.)

http://www.getcruise.com/

Eur

opea

n A

utos

January 26, 2015


18

OEMs – no clear leaders and hard to see how to build a sustainable advantage

If we look at the OEMs, it is hard to make out an investment case, as AVs will be such a small part of their operations for so long. A strong push into AVs by a firm like Volvo could “move the needle,” but there are no clear signs of the necessary major commitment by any OEM: as noted, they are pretty deeply conflicted about all this (see Toyota floor mats again). Google of course is a leader, and if indeed they can make the Google Car testbed really work (see my comments later), they may be a seller of technology to car companies, which might amount to significant revenue. In a different vein, if one believes that V2X (“connected car”) is going to be a big part of all this, then General Motors is the place to be, as this company has consistently over a long period been a big proponent of and investor in V2X technology. In a final tangential comment, if you feel that the whole LSV/NEV world is about to get an enormous boost from ADAS and AV engineering, then a close look at Polaris and Bombardier may make sense: these may be the new OEMs of the AV future.

9. Forecasting the future

Bernstein "Glenn, now that we have rather comprehensively surveyed the AV landscape, where do you ultimately think we think we are headed? Can you help us think about how fast this will happen and how to quantify what it means for everyone?"

Full Level 5 AVs at least a decade away

Glenn Mercer: "I’ll throw out a personal forecast which I’ll first summarize, and then lay out in more detail - we’ll capture 80% or more of the benefits of AVs without going to full level 5, which may be well over a decade away (in the mass market). Levels 1/2/3 will deliver most of the value in the near-term, with level 4 (essentially a switchable level 5) contributing more value in the longer-term, in controlled environments (e.g. low-speed full AV in small contained communities, some commercial delivery vehicles, high-speed highway AVs as a service that can be switched on to guide from exit to exit, under good conditions).

This view of the future, which is much more modest than the view of the most ardent supporters, is based on my sense that the risk factors will turn out to be bigger than feared. As mentioned earlier, if Toyota can be floored by some floor mats… If it took ABS almost 20 years and ACC (active cruise control) about 15 to get to 5% penetration in the USA, do we really think that the exponentially harder challenge of level 5AVs will happen faster?

Detail: Near term – driver assist to become common place

"I do see some low hanging fruit – by this I mean high likelihood of implementation, now and over the next decade. Obviously ADAS continues its steady penetration march, so that we back into AV levels 1/2/3, with lane warning, ACC, automated parking, traffic jam management, and more. This gets us 90% of the safety, and some of the convenience benefit, for a small portion of the level 4/5 cost, and more quickly.

Detail: Mid term – some AVs in controlled environments, possibly exit-to-exit highway management for cars

“By mid-term, I mean high likelihood, but slower: starting in serious volume in 5 years.

First, transit buses and “last mile” vehicles. This is a good place to try out full AVs, given low speeds, government involvement (government as operator and not just regulator), and the presence of traineddrivers on hand to monitor these experiments.

Eur

opea

n A

utos

January 26, 2015


19

In commercial vehicle applications, we can, as with transit, do more faster, given we have paid trained operators to handle the vehicles in the trials and experiments. And also truck fleets will adopt rapidly if theeconomics prove out (and we need to recall that commercial fleets are deeply skeptical of highly-touted theoretical advantages to many technologies, which when placed in real-world usage tend to fall short). Butlots of issues remain (e.g. regarding truck platooning, we have issues of wind generation, tailgating laws, interoperability across trucks OEMs, cooperation among fleets – and then researcher Alain Kornhauser has said that only 1 to 10% of truck miles traveled are platoonable anyway, which leads to very low fuel savings). Trucks will certainly adopt level 1/2/3 safety features with dispatch, especially if these help with HOS (hours of service) rules: will regulators allow drivers who have ADAS assistance to run longer shifts?

Low-speed shuttles will probably launch (sort of road-going AGVs (automatic guided vehicles)), in controlled environments like airport shuttle services, and possibly in urban taxi settings (“an Uber where I don’t have to talk to the driver.”)

In the realm of LSV/NEVs we should see some level 4 and even level 5 off-highway operations, e.g. for gated communities and local shopping trips or doctor visits. One early adopter might be a place like Babcock Ranch in Florida, (www.babcockranchflorida.com), a community designed from the outset with NEVs in mind.

For regular cars, probably some luxury vehicles will offer exit-to-exit highway management (a type of level 4 application), which turns off once we exit to surface streets, and would be limited to certain highways, certain weather conditions, maybe even certain lanes.

Are Glenn's views too conservative?

AV proponents would call this mid-term car forecast hopelessly conservative. This reflects to some extent a divide in the AV world between the high-speed and low-speed perspectives. There is a fundamental dichotomy between high-speed highway (defined as limited access highways) and low-speed surface street (regular urban and suburban streets, roads, some small highways) implementation of AVs. Some believe highway implementation comes first, as the situation is easier to model, given that all the vehicles are moving at roughly same speed, in the same direction, with no real intersections (just on and off ramps), no pedestrians to deal with, etc. But the failure mode of high-speed operation is horrible: deaths, fatalities. Others therefore see surface street implementation coming first: despite much greater complexity (numerous intersections of various kinds, unlimited access from cross streets and parking lots, pedestrians, bikes, crosswalks, mixes of directions and speeds and signage, etc.), the failure mode is friendlier: typically just bent metal or an injury, rather than a fatality. Taking everything into account, and trying to resolve the two perspectives, my view is we will see at high speeds only controlled-environment level 4 operations for a while (e.g. “on this highway in this weather for the next 10 exits you can switch to automated operation”), while at low speeds we will see some selected surface-road mobility experiments (“this AV, limited to 25 mph, will take this elderly citizen to the doctor and back”).

Note that Google’s pivot of research emphasis from high-speed highway operations (focused on convenience and safety) to low-speed surface street operations (focused on mobility) may reflect a similar thinking about the calculus of benefits and costs, especially failure mode. Google has said that in particular what was worrying them in highway tests was that human drivers were not holding up their end of the level 4 “bargain,” by not paying attention and being fully ready to resume control if needed. (Off the record, one informed source told me, “We’ll let the auto OEMs take on highway operations, we’ll focus on low-speed surface streets.”) See the Google Car discussion below.

Detail: Long term – when are where?

Ten years out, of course things are harder to forecast. But I think one’s long-term forecast will focus on the “level 3 to 4 chasm,” as mentioned earlier. The business case for migrating from level 0 to 1 to 2 to 3 is clear, straightforward, and manageable. Incrementally OEMs add ADAS features which bring more

http://www.babcockranchflorida.com/

Eur

opea

n A

utos

January 26, 2015


20

autonomy to the car, with customers paying incremental prices for each feature (e.g. “I’ll take advanced braking but skip the automatic parallel parking, as I live in the countryside”). Costs are accrued and investments made feature-by-feature. Ultimate control and thus liability lie with the driver. But when we leap across the chasm to level 4, costs soar (because, as noted, the car now must run all driving functions, not just selected tasks like lane management), careful incrementalism is replaced by a single bolder jump, and liability most likely transfers in large part or wholly to the OEM and its supply chain. (After all, if you market the car as driving itself, so that the driver can take a nap if he or she wishes, we can hardly then blame the driver for an accident that occurs while she or he is asleep!). OEMs such as BMW have in conference presentations echoed this view: it is a phase change between 3 and 4, like water boiling or ice freezing.

If we can figure a way across this chasm, then in ten years we may see a level 5 implementation, but I am not optimistic. Certainly we will expand the range of conditions under which level 4 can operate (right now, for example, there is not a viable level 4 car that can operate consistently in rain or snow). Maybe the closest thing to a level 5 (technically) will be an autonomous lorry crossing Australia, with little risk to anyone save kangaroos.

Where this will happen is entirely up for grabs. The US has so far allowed more on-road experimentation with AVs, but the litigation risk in America is huge, and state-by-state road conditions and regulations significant. Europe has better roads to deal with (e.g. more consistent lane markers and road signs for AVs to read, better road maintenance), but so far has seemed slow to embrace experiments. This might be (I am speculating) due to different cost/benefit analysis: Great Britain, for example, has only about 1,750 road fatalities annually, which of course is tragic, but is half the rate of fatalities from falls and about the same as fatalities by poison.

10. The Google Self-Driving Car

Bernstein: "Glenn, we've talked in detail about this technology, but for me most of the questions from investors are being inspired by Google's very public enthusiasm for the technology. I'm being told "Max, Google are going to put your old OEMs out of business!". What are your thoughts on this? Should I be worried?"

Glenn Mercer: "Given that Google is not only one of the leading proselytizers of AVs, but arguably one of the first firms to actually get one on the road (either retrofitted regular cars or the new purpose-built unit presumably being assembled – ironically enough, in Detroit – by Roush Industries), I agree we should talk a bit about its current effort.

Are Google backing away from a fully capable car already?

"In my view, this car represents a pivot (to use the Silicon Valley term) by Google away from a high-speed convenience value proposition, to a low-speed mobility proposition. Indeed, the initial official videos of the car (http://goo.gl/TXnhPn) prominently feature elderly people and even one blind person, tootling along on suburban streets, rather than Millennials cruising down the 101.

In my view, this makes good sense, both in terms of avoiding nasty failures (high-speed fatal crashes) and in addressing a huge market segment that to date has gotten little attention. While Google itself is unlikely

http://goo.gl/TXnhPn

Eur

opea

n A

utos

January 26, 2015


21

to build these cars, one can presume that they would be very interesting in selling the software and mapping system to any OEM interested."

Bernstein: "What does the Google Car tell us about the current state of the art, and likely trajectory, of AVs?"

Glenn Mercer: "Well, of course it is an amazing achievement, given that only a few years ago primitive AVs developed by university robotics departments were regularly crashing. And it or its predecessors have been tested over hundreds of thousands of miles, with very few incidents. Yet it does reveal just how much further we have to go.

As summarized by Lee Gomes (Slate, October 21, 2014) some of the car’s challenges are:

The Google car has travelled millions of miles…on the same roads, again and again

∑ Initial mapping effort. The car’s map (not by the way, an adaptation of Google Maps, which provides nowhere near the required level of detail), is very extensive, and thus very expensive, as it must contain exact locations (in three axes) of streetlights, stop signs, crosswalks, lane markings, and every other crucial aspect of a roadway, including curbstones. “In fact …Google admitted to me that the process it currently uses to make the maps is too inefficient to work in the country as a whole” (Gomes). To date, in fact, sensor scanner data is manually checked by humans to ensure accurate representation. The cars may have travelled many miles, but to date these are the same few thousand miles, mapped and remapped. As yet untouched are the 4 million other miles of US public roads.

Maps may not be able to keep up with constant small changes to the road network

∑ Map updating effort. Of course the car’s map will never be final, as it must be updated for running changes on the roadways. This also will be difficult, given how many traffic signs, signals, lane markings, etc. get added or moved or deleted daily in the USA (typically without logging the changes into any database, either promptly or at all). For example, the car may or may not stop at a traffic signal that is not in its map, and as we all know, transportation departments put up and take down temporary signals every day. And here is where the law of large numbers works against AVs: with 3 trillion miles driven every year in the USA, the rare event becomes the common event. As MIT roboticist John Leonard has explained, while the probability of any particular driver encountering a changed signal or sign may low, the probability of at least one driver seeing such on any given day is very high.

It is incredibly card to transfer human common sense to lines of code

∑ Processing effort. The car’s computers will of course work diligently to deal with all sorts of road events that humans take as second nature. But many of these are just incredibly difficult to transfer from generalized human “common sense” to lines of code. Leonard and others have accumulated lists of these situations: the policeman in the road who is waving you on, the head nod from a driver in a car letting you into a line of traffic, the crumpled paper bag in the road that looks like a rock to lidar, the child’s ball rolling into the street, or even the child who on Halloween steps into the road dressed as a ghost, in a white sheet of cloth. So much of this is so easy for human drivers (and driving is easy for us, given that there are several hundred million licensed drivers on the planet), but so hard to build into machine intelligence. Will the task be accomplished? I think it will – but at a slower pace than the enthusiasts expect.

∑ Judgment effort. One should consult the work of Bryant Walker Smith, in my view America’s leading scholar on the legal and ethical issues of AVs, for much more detail here, but I’ll tee up the subject with just one thought experiment, as laid out in an AV conference small-group discussion session. You are parked at a regular crossroads. The traffic light is red. In your rearview mirror you see an out-of-control car careening toward you at high speed. A violent accident is imminent. You, as a human, look both ways down the cross street, see no cars, and “floor it” across the intersection, running the red light and

Eur

opea

n A

utos

January 26, 2015


22

averting the collision, avoiding injury or even death. Now, put the AV in charge. How do we program it to know when to break the law, and run the red light? We may all need to dust off Isaac Asimov’s Laws of Robotics, first laid out in 1942.

Eur

opea

n A

utos

January 26, 2015


23

Disclosure Appendix

Ticker Table

Ticker Rating CUR

22 Jan 2015ClosingPrice

TargetPrice

TTMRel.Perf.

EPS P/E

2013A 2014E 2015E 2013A 2014E 2015E Yield

BMW.GR M EUR 98.65 100.00 17.7% 8.10 8.76 9.31 12.2 11.3 10.6 2.6%

DAI.GR M EUR 77.59 65.00 19.3% 6.40 7.60 6.78 12.1 10.2 11.4 2.9%

FCA.IM U EUR 11.04 6.25 -2.7% 0.74 0.19 0.75 14.8 59.4 14.8 NA

PAH3.GR O EUR 73.49 100.00 -6.8% 7.86 10.56 11.72 9.3 7.0 6.3 2.7%

RNO.FP M EUR 67.52 75.00 1.5% 2.15 6.09 9.10 31.4 11.1 7.4 2.6%

UG.FP O EUR 12.09 15.00 29.4% -6.77 -0.39 1.05 NM NM 11.5 NA

VOW.GR O EUR 194.85 260.00 0.7% 18.63 22.89 24.82 10.5 8.5 7.8 2.1%

VOW3.GR O EUR 198.40 260.00 -0.7% 18.63 22.89 24.82 10.6 8.7 8.0 2.1%

1114.HK (Brilliance) O HKD 14.48 17.00 14.2% 0.67 1.00 1.21 17.3 11.6 9.6 0.7%

489.HK (Dongfeng) M HKD 11.80 15.00 -1.4% 1.22 1.52 1.62 7.7 6.2 5.8 1.6%

2333.HK (Great Wall) U HKD 43.10 30.00 14.5% 2.70 2.47 2.98 12.8 14.0 11.6 1.7%

2238.HK (GAC) M HKD 7.21 8.00 -19.6% 0.40 0.67 0.79 14.3 8.6 7.3 0.0%

175.HK (Geely) U HKD 3.12 2.00 -10.4% 0.32 0.25 0.29 7.9 10.2 8.7 1.5%

7201.JP (Nissan) O JPY 1035.00 1200.00 5.9% 92.82 117.29 133.76 11.2 8.8 7.7 2.4%

TTMT.IN M INR 566.50 530.00 47.5% 43.51 60.67 74.56 13.0 9.3 7.6 0.4%

TTM M USD 48.84 44.17 47.5% 3.63 5.06 6.21 13.5 9.7 7.9 0.3%

200625.CH (Changan-B) O HKD 19.40 28.50 30.1% 0.75 1.57 2.37 20.7 9.9 6.6 0.5%

000625.CH (Changan) O CNY 19.88 23.00 74.4% 0.75 1.57 2.37 26.5 12.7 8.4 0.5%

600104.CH (SAIC) O CNY 23.79 29.00 77.1% 2.25 2.50 3.09 10.6 9.5 7.7 5.0%

MSDLE15 1417.43 87.14 89.75 98.21 16.3 15.8 14.4 3.3%

SPX 2063.15 108.28 116.04 125.22 19.1 17.8 16.5 2.0%

MXAPJ 476.32 32.96 35.34 38.46 14.5 13.5 12.4 3.1%

O – Outperform, M – Market-Perform, U – Underperform, N – Not Rated

TTMT.IN estimates are 2014A/2015E/2016E

Valuation Methodology

In a "normal" economy, we value stocks on the basis of their returns on capital, using the calculation "EV/IC = (ROIC – g)/(WACC – g)" where g accounts for growth. We believe that over time, even in the momentum-driven auto sector, valuations will be driven by the ability of a company to generate a return on its capital base and grow its business. In a "normal" economy, we also look at EV/EBITDA and P/E to gauge relative valuations and peak stock price potential. In more challenging times, when earnings are minimal and stocks de-rate, we also look at valuations versus historical troughs on metrics such as EV/IC and EV/sales and look at balance sheet strength.

Risks

The risks to our view on Chinese auto stocks and our share price targets are straightforward and are mainly macroeconomic in nature. Earnings, liquidity and equity value could be severely tested in the event of an economic slowdown in China. The individual companies are at risk of specific product and project failure. The highly politicized nature of the Chinese auto industry creates a number of related risks, both external (e.g., Japan protests) and internal (e.g., Chinese government intervention in policy or companies). All the companies are audited by international audit firms and appear to use conventional accounting. However, there appear to be odd working capital practices, unclear inter-company relationships and politicized corporate governance. Currency is a limited risk for the stocks.

The risks to our view on European Autos stocks and our share price targets are straightforward and are mainly macroeconomic in nature. Earnings, liquidity and equity value could be severely tested in the event

Eur

opea

n A

utos

January 26, 2015


24

of a double-dip recession proves even deeper and longer than we forecast and if auto and truck sales fall far below our assumptions, putting our price targets for the European Autos stocks at risk. The ability of the financial services businesses to remain viable is at risk if the financial system deteriorates again and capital market access becomes impossible.

Risks specific to Tata include execution risk associated with JLR's rapid expansion, and the ongoing weakness of the Indian standalone business which could deteriorate further in a highly competitive market environment. Our forecasts for Tata are also sensitive to moves in the Pound versus the US Dollar, Euro, Chinese RMB, and the Indian Rupee.

Our forecasts are also sensitive to moves in the Euro versus the US dollar and the UK sterling as well as Latin American and Asian currencies.

SRO REQUIRED DISCLOSURES

∑ References to "Bernstein" relate to Sanford C. Bernstein & Co., LLC, Sanford C. Bernstein Limited, Sanford C. Bernstein (Hong Kong) Limited 盛博香港有限公司, and Sanford C. Bernstein (business registration number 53193989L), a unit of AllianceBernstein (Singapore) Ltd. which is a licensed entity under the Securities and Futures Act and registered with Company Registration No. 199703364C, collectively.

∑ Bernstein analysts are compensated based on aggregate contributions to the research franchise as measured by account penetration, productivity and proactivity of investment ideas. No analysts are compensated based on performance in, or contributions to, generating investment banking revenues.

∑ Bernstein rates stocks based on forecasts of relative performance for the next 6-12 months versus the S&P 500 for stocks listed on the U.S. and Canadian exchanges, versus the MSCI Pan Europe Index for stocks listed on the European exchanges (except for Russiancompanies), versus the MSCI Emerging Markets Index for Russian companies and stocks listed on emerging markets exchanges outside of the Asia Pacific region, and versus the MSCI Asia Pacific ex-Japan Index for stocks listed on the Asian (ex-Japan) exchanges - unless otherwise specified. We have three categories of ratings:

Outperform: Stock will outpace the market index by more than 15 pp in the year ahead.

Market-Perform: Stock will perform in line with the market index to within +/-15 pp in the year ahead.

Underperform: Stock will trail the performance of the market index by more than 15 pp in the year ahead.

Not Rated: The stock Rating, Target Price and estimates (if any) have been suspended temporarily.

∑ As of 01/22/2015, Bernstein's ratings were distributed as follows: Outperform - 47.0% (2.1% banking clients) ; Market-Perform - 41.4% (1.6% banking clients); Underperform - 11.6% (0.0% banking clients); Not Rated - 0.0% (0.0% banking clients). The numbers in parentheses represent the percentage of companies in each category to whom Bernstein provided investment banking services within the last twelve (12) months.

∑ Accounts over which Bernstein and/or their affiliates exercise investment discretion own more than 1% of the outstanding common stock of the following companies 2333.HK / Great Wall Motor Co Ltd.

∑ This research publication covers six or more companies. For price chart disclosures, please visit www.bernsteinresearch.com, you can also write to either: Sanford C. Bernstein & Co. LLC, Director of Compliance, 1345 Avenue of the Americas, New York, N.Y. 10105 or Sanford C. Bernstein Limited, Director of Compliance, 50 Berkeley Street, London W1J 8SB, United Kingdom; or Sanford C. Bernstein (Hong Kong) Limited 盛博香港有限公司, Director of Compliance, Suites 3206-11, 32/F, One International Finance Centre, 1 Harbour View Street, Central, Hong Kong, or Sanford C. Bernstein (business registration number 53193989L) , a unit of AllianceBernstein (Singapore) Ltd. which is a licensed entity under the Securities and Futures Act and registered with Company Registration No. 199703364C, Director of Compliance, 30 Cecil Street, #28-08 Prudential Tower, Singapore 049712.

12-Month Rating History as of 01/22/2015

Ticker Rating Changes

000625.CH O (IC) 10/20/14

1114.HK O (IC) 02/19/13

175.HK U (IC) 02/19/13

200625.CH O (IC) 10/20/14

2238.HK M (IC) 02/19/13

2333.HK U (RC) 11/13/14 M (IC) 02/19/13

489.HK M (RC) 08/01/14 O (IC) 02/19/13

600104.CH O (IC) 10/14/14

7201.JP O (RC) 01/13/14

BMW.GR M (RC) 04/09/14 O (RC) 03/23/11

DAI.GR M (RC) 03/16/12

FCA.IM U (IC) 11/11/14

PAH3.GR O (RC) 04/09/14 M (RC) 07/23/12

RNO.FP M (RC) 05/29/14 O (RC) 01/13/14

TTM M (RC) 09/03/14 O (IC) 09/27/13

TTMT.IN M (RC) 09/03/14 O (IC) 09/24/13

UG.FP O (RC) 01/13/14

VOW.GR O (RC) 04/09/14 M (RC) 08/17/10

VOW3.GR O (RC) 04/09/14 M (RC) 08/17/10

Rating Guide: O - Outperform, M - Market-Perform, U - Underperform, N - Not Rated

Rating Actions: IC - Initiated Coverage, DC - Dropped Coverage, RC - Rating Change

OTHER DISCLOSURES

A price movement of a security which may be temporary will not necessarily trigger a recommendation change. Bernstein will advise as and when coverage of securities commences and ceases. Bernstein has no policy or standard as to the frequency of any updates or changes to its coverage policies. Although the definition and application of these methods are based on generally accepted industry practices and models, please note that there is a range of reasonable variations within these models. The application of models typically depends on forecasts of a range of economic variables, which may include, but not limited to, interest rates, exchange rates, earnings, cash flows and risk factors that are subject to uncertainty and also may change over time. Any valuation is dependent upon the subjective opinion of the analysts carrying out this valuation.

This document may not be passed on to any person in the United Kingdom (i) who is a retail client (ii) unless that person or entity qualifies as an authorised person or exempt person within the meaning of section 19 of the UK Financial Services and Markets Act 2000 (the "Act"), or qualifies as a person to whom the financial promotion restriction imposed by the Act does not apply by virtue of the Financial Services and Markets Act 2000 (Financial Promotion) Order 2005, or is a person classified as an "professional client" for the purposes of the Conduct of Business Rules of the Financial Conduct Authority.

To our readers in the United States: Sanford C. Bernstein & Co., LLC is distributing this publication in the United States and accepts responsibility for its contents. Any U.S. person receiving this publication and wishing to effect securities transactions in any security discussed herein should do so only through Sanford C. Bernstein & Co., LLC.

To our readers in the United Kingdom: This publication has been issued or approved for issue in the United Kingdom by Sanford C. Bernstein Limited, authorised and regulated by the Financial Conduct Authority and located at 50 Berkeley Street, London W1J 8SB, +44 (0)20-7170-5000.

To our readers in member states of the EEA: This publication is being distributed in the EEA by Sanford C. Bernstein Limited, which is authorised and regulated in the United Kingdom by the Financial Conduct Authority and holds a passport under the Markets in Financial Instruments Directive.

To our readers in Hong Kong: This publication is being distributed in Hong Kong by Sanford C. Bernstein (Hong Kong) Limited 盛博香港有限公司, which is licensed and regulated by the Hong Kong Securities and Futures Commission (Central Entity No. AXC846). This publication is solely for professional investors only, as defined in the Securities and Futures Ordinance (Cap. 571).

To our readers in Singapore: This publication is being distributed in Singapore by Sanford C. Bernstein, a unit of AllianceBernstein (Singapore) Ltd., only to accredited investors or institutional investors, as defined in the Securities and Futures Act (Chapter 289). Recipients in Singapore should contact AllianceBernstein (Singapore) Ltd. in respect of matters arising from, or in connection with, this publication. AllianceBernstein (Singapore) Ltd. is a licensed entity under the Securities and Futures Act and registered with Company Registration No. 199703364C. It is regulated by the Monetary Authority of Singapore and located at 30 Cecil Street, #28-08 Prudential Tower, Singapore 049712, +65-62304600. The business name "Sanford C. Bernstein" is registered under business registration number 53193989L.

To our readers in the People’s Republic of China: The securities referred to in this document are not being offered or sold and may not be offered or sold, directly or indirectly, in the People's Republic of China (for such purposes, not including the Hong Kong and Macau Special Administrative Regions or Taiwan), except as permitted by the securities laws of the People’s Republic of China.

To our readers in Japan: This document is not delivered to you for marketing purposes, and any information provided herein should not be construed as a recommendation, solicitation or offer to buy or sell any securities or related financial products.

To our readers in Australia: Sanford C. Bernstein & Co., LLC, Sanford C. Bernstein Limited and Sanford C. Bernstein (Hong Kong) Limited 盛博香港有限公司 are exempt from the requirement to hold an Australian financial services licence under the Corporations Act 2001 in respect of the provision of the following financial services to wholesale clients:

∑ providing financial product advice;

∑ dealing in a financial product;

∑ making a market for a financial product; and

∑ providing a custodial or depository service.

Sanford C. Bernstein & Co., LLC., Sanford C. Bernstein Limited, Sanford C. Bernstein (Hong Kong) Limited 盛博香港有限公司 and AllianceBernstein (Singapore) Ltd. are regulated by, respectively, the Securities and Exchange Commission under U.S. laws, by the Financial Conduct Authority under U.K. laws, by the Hong Kong Securities and Futures Commission under Hong Kong laws, and by the Monetary Authority of Singapore under Singapore laws, all of which differ from Australian laws.

One or more of the officers, directors, or employees of Sanford C. Bernstein & Co., LLC, Sanford C. Bernstein Limited, Sanford C. Bernstein (Hong Kong) Limited 盛博香港有限公司, Sanford C. Bernstein (business registration number 53193989L) , a unit of AllianceBernstein (Singapore) Ltd. which is a licensed entity under the Securities and Futures Act and registered with Company Registration No. 199703364C, and/or their affiliates may at any time hold, increase or decrease positions in securities of any company mentioned herein.

Bernstein or its affiliates may provide investment management or other services to the pension or profit sharing plans, or employees of any company mentioned herein, and may give advice to others as to investments in such companies. These entities may effect transactions that are similar to or different from those recommended herein.

Bernstein Research Publications are disseminated to our customers through posting on the firm's password protected website, www.bernsteinresearch.com. Additionally, Bernstein Research Publications are available through email, postal mail and commercial research portals. If you wish to alter your current distribution method, please contact your salesperson for details.

Bernstein and/or its affiliates do and seek to do business with companies covered in its research publications. As a result, investors should be aware that Bernstein and/or its affiliates may have a conflict of interest that could affect the objectivity of this publication. Investors shouldconsider this publication as only a single factor in making their investment decisions.

This publication has been published and distributed in accordance with Bernstein's policy for management of conflicts of interest in investment research, a copy of which is available from Sanford C. Bernstein & Co., LLC, Director of Compliance, 1345 Avenue of the Americas, New York, N.Y. 10105, Sanford C. Bernstein Limited, Director of Compliance, 50 Berkeley Street, London W1J 8SB, United Kingdom, or Sanford C. Bernstein (Hong Kong) Limited 盛博香港有限公司, Director of Compliance, Suites 3206-11, 32/F, One International Finance Centre, 1 Harbour View Street, Central, Hong Kong, or Sanford C. Bernstein (business registration number 53193989L) , a unit of AllianceBernstein (Singapore) Ltd. which is a licensed entity under the Securities and Futures Act and registered with Company Registration No. 199703364C, Director of Compliance, 30 Cecil Street, #28-08 Prudential Tower, Singapore 049712. Additional disclosures and information regarding Bernstein's business are available on our website www.bernsteinresearch.com.

CERTIFICATIONS

∑ I/(we), Max Warburton, Robin Zhu, Senior Analyst(s)/Analyst(s), certify that all of the views expressed in this publication accurately reflect my/(our) personal views about any and all of the subject securities or issuers and that no part of my/(our) compensation was, is, or will be, directly or indirectly, related to the specific recommendations or views in this publication.

Approved By: CDK

Copyright 2015, Sanford C. Bernstein & Co., LLC, Sanford C. Bernstein Limited, Sanford C. Bernstein (Hong Kong) Limited 盛博香港有限公司, and AllianceBernstein (Singapore) Ltd., subsidiaries of AllianceBernstein L.P. ~1345 Avenue of the Americas ~ NY, NY 10105 ~212/756-4400. All rights reserved.

This publication is not directed to, or intended for distribution to or use by, any person or entity who is a citizen or resident of, or located in any locality, state, country or other jurisdiction where such distribution, publication, availability or use would be contrary to law or regulation or which would subject Bernstein or any of their subsidiaries or affiliates to any registration or licensing requirement within such jurisdiction. This publication is based upon public sources we believe to be reliable, but no representation is made by us that the publication is accurate or complete. We do not undertake to advise you of any change in the reported information or in the opinions herein. This publication was prepared and issued by Bernstein for distribution to eligible counterparties or professional clients. This publication is not an offer to buy or sell any security, and it does not constitute investment, legal or tax advice. The investments referred to herein may not be suitable for you. Investors must make their own investment decisions in consultation with their professional advisors in light of their specific circumstances. The value of investments may fluctuate, and investments that are denominated in foreign currencies may fluctuate in value as a result of exposure to exchange rate movements. Information about past performance of an investment is not necessarily a guide to, indicator of, or assurance of, future performance.

Documents

Autonomous Vehicles