The Race For The Auto Cockpit: Nvidia, Qualcomm Or Mobileye? (NASDAQ:NVDA)

Secular Semiconductor Auto Sales Growth

Amid the ChatGPT and “Artificial Intelligence” AI excitement and hype, the semiconductor auto segment seems to have been forgotten and relegated to an afterthought. This is a mistake. I believe this segment will see outstanding growth and is worth focusing on. Nvidia Corporation (NASDAQ:NVDA), Qualcomm Incorporated (NASDAQ:QCOM), and Mobileye Global Inc. (MBLY) are the three big guns in this sector, and we should take a closer look at their future prospects.

As we can see above, all three have large pipelines of $14Bn, $22Bn and $30Bn, respectively, totaling $66Bn. They have estimated auto sales of $5.4Bn in 2023; that leaves over $60Bn of the pipeline promising secular growth for the next 5 to 7 years.

Mobileye, an Intel Corporation (INTC) subsidiary, has the highest estimated sales at $2.1Mn in 2023, with Qualcomm snapping at its heels at $1.95Bn. Mobileye’s first quarter growth was only 16%, slowing down after 2022’s heady growth of 35%; but I expect it grow at a CAGR of 39% in the next 3 years. Mobileye is already in 233 models world-wide and is the standard bearer when it comes to vision hardware and software.

I had earlier recommended Qualcomm in March 2023, asserting that Auto’s would make up for the loss of Apple’s I-Phone business over time. After a high 41% growth in 2022, Qualcomm had a slower March 2023 quarter growth of only 20%, but will continue to grow at a CAGR of 32% in the next 3 years. Qualcomm’s chief strength is its connectivity, and its focus is to be the seamless, one stop shop integrating ADAS, Infotainment, Car to Cloud services and connectivity.

Big Daddy Nvidia, now riding the “Generational AI” inflection wave, grew its auto segment the fastest among the three in the previous year at 69%, on a smaller base and is likely to grow 45% this year, with a strong CAGR of 40% in the next three years. Nvidia’s focus is being the agnostic all-purpose supplier for hardware, software and cloud services. It positions itself as the datacenter on wheels, and leverages all its existing strengths in data center, omniverse and high speed computing.

A Computer On Wheels – Growth And Trends

As Elon Musk predicted in 2015

We really designed the Model S to be a very sophisticated computer on wheels

There are varying estimates for the global semiconductor market size and a range of growth estimates.

Reportlinker believes it would grow from $59Bn in 2022 at a CAGR of 8.4% to $104Bn, while Fortunebusinessinsights.com thinks it will grow at a faster CAGR of 12.6% during the same period. However, Fortune pegs the market size much smaller, only reaching $73Bn by 2028, implying a current market size of $36Bn. Nonetheless, regardless of the methodology used for sizing, growth rates ranging between 8 to 12% is great. More importantly, the three companies’ pipelines and revenue growth estimates suggest much stronger growth than these two industry studies.

As we can see below from the end user segments, Auto’s share of global semiconductor revenue was only 14% in 2022, but it has the highest growth prospects at 14.1%

Semiconductor usage grows with the increased use of automation for:

• Supporting Vehicle Artificial Intelligence.
• Advancing through Levels 1 (Driver Assistance) to Level 5 (Full Autonomy)
• ADAS (Advanced Driver Assist Systems) for safer driving.
• Digital connectivity.
• The adoption of hybrid and electric vehicles, which require a greater proportion of semiconductors, with conventional vehicles requiring $300 per car and hybrid and electric vehicles needing $1,000.
• Integrated platforms supporting all vehicle stacks.
• The growth of Robotaxis and the march towards Fully Autonomous Vehicles.

Strategy And Product Differentiators

Mobileye

Strong Focus on Vision – Mobileye as the name suggests is the strongest in vision with the bulk of its sales coming from the EyeQ System-on-Chip (SoC), which is the primary vision assist for ADAS The EyeQ system consists of cameras and sensors to enable drivers to stay in their lane, spot blind spots and avoid collisions by providing alerts and emergency braking.

Incremental Move from ADAS to Full Autonomy – Mobileye believes that full autonomy is at least 25 years away and will not be achieved at scale till at least 2050. Their focus has always been incremental as seen below and growing from ADAS would be the better, practical, economical, and profitable way to achieve full autonomy. Their focus has also been on Robotaxis, which is currently deployed at a higher scale than even the most advanced Autonomous Vehicle.

As we can from the two slides below, Mobileye’s incremental approach is a rational and cost-effective way to advance, currently we are only at level 2, and level 3 is not even legal in the U.S., while Robotaxis are in just a few select areas. They do have the full product portfolio to get there.

Mobileye’s strength of an integrated platform, which provides clients all levels of vision and other support is great and often preferred. They are after all the market leader. However, it tends to act as a walled garden sometimes and integrating it with other stacks can be a deterrent if quality perceptions differ. Nvidia is the easiest building block, while Qualcomm also offers flexibility, but Mobileye will more likely prefer clients build up upon existing Mobileye products.

Nvidia

For readers wanting to do a deep dive into Nvidia’s strategy there is an excellent Q&A worth listening to between New Street Research and Nvidia’s VP and General Manager of Auto.

Nvidia’s key strategies and main strengths in auto are:

Being an End-to-End Platform for partners, providing all solutions including

• Centralized operating systems for all stacks including Cockpit, EV, Safety, DMS/OMS/CMS (Driver, Occupant, and Crash Management Safety)
• Car design.
• Artificial Intelligence model training
• Simulated data in addition to customers’ own data.
• Automated factory design using Omniverse synthetic data generation.
• Best in class high performance hardware.
• Modular or full stack, hardware, software, or cloud, Nvidia is more focused on being everyone’s partner and building relationships.

Positioning Thor as the “Superchip”

Consistent with its positioning as the best in its class SoC’s in datacenter, gaming and Omniverse, Nvidia touts Thor as the fastest, high performance superchip in the Thor central platform.

Increase the user base: Their prime goal is enhancing their user base in all aspects of auto digitization and from their ecosystem, it’s clear that they are everywhere. For example, they are in 8 of the largest Robotaxis and their concierge solutions are for level 4 and level 5 for Mercedes and Jaguar Land Rover.

Another good example of their strategy is their partnership with Mediatek, to continue making inroads into the auto sector even if the product/service is sold under the Mediatek brand as MediaTek SoC for auto but has Nvidia’s chips inside.

High Performance Computing Cloud – Datacenter on wheels.

Leverage their existing datacenter strengths by providing simulation and data center usage to all customers, including Tesla, which has its own SoC’s but uses Nvidia for datacenter.

Move up the value chain.

They have a great deal of flexibility with their OEMS and other partners, knowing that eventually everyone wants to control their own software all the way to Levels 3,4 and 5, Nvidia believes their platform is wide enough and will benefit everyone who wants to hook into their ecosystem. Nvidia believes that Fully Autonomous Vehicles are at least 15 years away.

Network Effect

This is a little bit like a network effect because they have so many different partners, and so many different customers, and the data off all the customers of this data can be added, anonymized, and used by everyone in simulations.

Flexibility

Because requirements and networks are growing, customers need a general-purpose programmable architecture because they don’t know what algorithms they’ll be needing in the future. Customers and partners should be able to work on the same CUDA, APIs, and tensor architecture, which is compatible across generations, therefore it can easily increase and enhance performance across older software.

The main point with digitizing the car is that even today, maybe a decade and a half in, there are already a very wide and disparate network of computers inside a car, which often have trouble talking to one another. Nvidia’s strength is to bring it all together.

The slide above shows Nvidia’s main pipeline customers and ones on the right have the full stack, Jaguar Land Rover, and Mercedes. They do have a very strong presence in the Chinese market, with several automakers and OEMs in their fold with big ramp-ups from XPeng (XPEV) and NIO Inc. (NIO).

Like everyone else they too, would prefer high value recurring software revenue from L4 and L5, full stack solutions, but right now they are in the middle. Their strategy is to provide modular solutions with the endgame of letting customers use their platform to move up the value chain. They realize that full autonomy is more than a decade away and they also believe that their platform will contribute significantly to it. Therefore, lower cost customers can take the modules they can afford and continue working with their other vendors if they choose to do so. Several customers and partners do not have the financial or tech capacity to evolve and Nvidia’s strategy is work with them to realize that capacity.

Nvidia emphasizes that they have a large investment in their platform and would like to leverage all their competitive advantages of datacenter, omniverse and high-performance computing and their main goal is to increase their installed base as they see scope for further revenues down the road, unlike other competitors like Qualcomm, who are not likely to go below Level 2, or Mobileye that would like to keep its black box closed.

Qualcomm

Key takeaways on Qualcomm’s auto strategy and strengths from a very interesting Q&A with VP Nakul Dugal VP Auto, Qualcomm.

Qualcomm wants to be your “Digital Chassis.”

Building a services platform at scale – ICE automakers must move to a better platform and Qualcomm understands a new architecture will be needed at scale with product, connectivity, and services. Qualcomm would be the best to deal with the complexity because it would be fully integrated, i.e., they don’t want to dabble in components, instead they would be the digital chassis. Scale is important, Qualcomm is launching 150 programs in the next six quarters, each program entails working closely with an auto partner from design to production for 18 to 30 months.

Recurring software and platform revenues on the anvil at a later stage – This is an excellent long-term strategy because the life of the platform is so much longer and thus much more value can be extracted instead of by selling components, by both the automaker and Qualcomm. The business model is to sell 50 to 60% of the capability with scalable hardware with the ability to add software at later stages, as compliance needs increase. Additionally, customers could decide by functionality and keep adding as their needs increase on the same platform without adding different platforms.

Upselling is key – The goal 10-20 years out is to continuously upsell as a SaaS or PaaS business; Over time car makers would have a revenue stream either under their own banner or in partnership with Qualcomm. I believe Qualcomm as the tech company will have the larger or at least equal share in upsells, because configurations will be constantly evolving. Pay when you upgrade.

The smartphone experience is a big competitive advantage and Qualcomm is utilizing it to the hilt in design, connectivity, scalability. Also, important has been their experience navigating environments form 2G to 5G, which will be a key component of autos as we move from L1 to L5. That synergy is invaluable and its strength that has led to their adoption in so many platforms. Complexity makes Qualcomm relevant and needed to manage different systems, stacks, vendors.

Quoting Nakul Dugal, VP, Auto Qualcomm, emphasis mine:

And the concurrencies are extremely complex, because you are serving three customers, you’re serving the car, you are serving the real-time nature of how you are engaging with the driver, and you are serving an external application ecosystem that is coming in. So, we’ve been at this for now close to a decade and the one big advantage that we had was that we were able to look at our smartphone roadmap, pick the right parts, make the right changes, and make available very quickly, solutions that serve the needs of every tier. And over time, we were able to influence the direction of how you actually design this cockpit, because it is really about integration, reducing the role of cost of ownership bringing down — bringing more functionality to lower tiers of the vehicle. And that has made us obviously, very strong platform provider. And I think first, I think there’s a lot of experience. We have kind of taken this market in the direction that it is gone. You have to be able to support this globally because automakers are all developing their solutions all over the world. So, you have to be able to have teams that can actually make that happen. You have to work with a large number of ecosystem partners because a cockpit is built by between a dozen and two dozen different software providers that come together. You have to build a roadmap that scales. So, we had to go acquire a team that had that experience. And the path that we are now on is, that it’s a flexible roadmap, we can host other people’s stacks, we’re obviously building our own.

Differentiating with Mobileye: Qualcomm sees Mobileye as their largest competitor and believes that their proprietary approach of selling their entire platform or a Blackbox is a big challenge for automakers who would feel the need to program their systems differently. Qualcomm could take market share being more flexible and building heterogeneous SoC’s the way they did in phones and IoT.

Connectivity is Qualcomm’s key strength, given their decades of experience in modems and IoT and they emphasize that even though they would prefer a walled platform of their own, they are open to unbundling for non-connectivity products.

Investment Case

While this article is about Nvidia, Qualcomm and Mobileye, the semiconductor auto space will also be dominated by Tesla, Inc. (TSLA); With its almost 2 decades in making electric cars it has the most end to end experience in digitization and the capacity to scale. It has already captured the charging stations market, and over time will be selling cockpit software to other automakers. Selling 1.8Mn vehicles this year with about $1,000 worth of silicon per car ($1.8Bn), makes it a very sizable presence in this industry. But that’s the scope of another article…I include it here to emphasize that Tesla would definitely be at the forefront of change in the auto industry.

Paradigm shift: ICE (Internal Combustion Engine) automakers have to make a paradigm shift embracing technology as the main driver and not looking at digitization partners as parts suppliers. Tesla, Nvidia, Qualcomm and Mobileye are demonstrating that technology at scale is really the only way to go and unless ICE or traditional auto companies don’t take a paradigm shift, they will be left behind. For example, a BMW 7 Series used to have like 150 different ECUs (Electronic Control Unit) five or six years ago. Today, it probably has more like 10.

I also believe – In a sense, there could be an inflection point for the auto segment as well, the way Jensen Huang, visualizes a revamp of all data centers. With the advent of Superchips like the H100 we could reach that critical mass where the next generation of transport would mean a complete revamp of the assembly line and the shop floor. Twenty years down the road, either these four would have completely changed the auto industry or would have been supplanted by an even nimbler upstart from Silicon Valley or China.

Fragmented Industry: The industry is fragmented with tremendous amount of competition, especially with electrification – the barriers to entry to build an EV has actually gone down drastically. In China a lot of newcomers from the tech and software space don’t really have any background in building cars, but because there is a tremendous amount of infrastructure and government support available to get into the market, they have reduced the barriers to building a software-defined vehicle.

This is great for suppliers who can provide teams, hardware, technology, capital, and scalability to cater to an always changing market, from compliance requirements and to the needs of various consumers. This could develop into an oligopoly for these four in the next decade.

A fragmented customer base may reduce customers’ bargaining power, but on the other hand, suppliers like Qualcomm and Nvidia have to produce at scale and be nimble at the same time to satisfy the customer base.

A Gold Rush? There is an element of a gold rush in this fragmented industry, especially in China, which like the U.S. is at the forefront because of government support. Who better than the providers of scalable picks and shovels to benefit?

A growing and fragmented market should help all four competitors.

I believe each vendor’s distinctive competency is needed to satisfy the myriad components of this fragmented market — with a large number of players (big and small) in two continents and two industries, Auto and Technology, still finding their way.

Scale, integration, wide offerings, flexibility, and ability to partner would be key.

A fragmented customer base is always better for suppliers and with both U.S. and Chinese governments throwing money at this sector I can see all four doing well for the next decade.

However, cyclical demand from autos is a key factor and regardless of the pace of digitization/electrification, this could be an enormous challenge because of the R&D and Capex needed for constant product and AI improvements.

The computerized vehicle is the one that really needs AI to succeed. Getting from Level 2 to full autonomy is not possible without high performance computing, AI and machine learning. There is going to be so much need for data generation, to reach levels 3,4 and 5 – to substitute the human brain, you need in an insane amount of computing power. AI needs to evolve and focus on building safer transportation instead of fixing high school essays.

This is a sector worth buying into for the next decade.

I own Nvidia, Qualcomm and Tesla. I had recommended Nvidia as a Buy in March 2023 and last bought more in May. I plan to hold it for a decade, it’s too large a part of my portfolio to buy more. I have a buy recommendation on Qualcomm, mainly because of auto, and continue to add it. Mobileye is a hold for now, but merits a closer look on declines.