From PCs to cars: Nvidia, Qualcomm and Intel race to automotive semis

From PCs to cars: Nvidia, Qualcomm and Intel race to automotive semis

Forward-looking: One of the biggest challenges for large semiconductor companies is that they must make big bets on key technologies and markets many years before they become commonplace. In our increasingly digital world, chips are the main base-level component necessary to enable new types of innovations that, in turn, lead to the creation of new product categories or the reinvention of older ones. Such is the case of today’s automobile market, which is seeing a transformation of capabilities, architecture, and design all being driven by chip-powered digital technologies.

Several years back, many of the world’s largest semiconductor companies, including Nvidia, Qualcomm, Intel and others started making plans for what they saw as the next big digital device category: the smart connected car.

Now, as we start to see the debut of model year 2023 cars with increasingly sophisticated tech-enabled functions, it’s becoming clear that those early bets are starting to payoff. Each of these major tech vendors — all of whom have typically focused on end user computing devices like PCs and smartphones — are making important inroads into the automotive industry. In fact, each are now starting to highlight signs of the increasing importance that being a car component supplier is having on their respective bottom lines.

At its recent GTC conference, for example, Nvidia highlighted how it is leveraging its next generation GPU technology for a variety of different automotive applications, including assisted and autonomous driving, as well as infotainment functions. Similarly, Qualcomm held an investor day last month that discussed the company’s rapidly growing pipeline of automotive-based business. Finally, Intel just announced that it will be spinning out its Mobileye division in what it hopes will prove to be a big financial windfall after purchasing the ADAS (Advanced Driver-Assistance Systems) technology-focused, Israel-based company in 2017.

In Nvidia’s case, the company has already released many previous iterations of its GPU-powered technology for the car market, but with each successive generation, the company is enabling more powerful capabilities and reaching into new applications. After years of success driving car infotainment systems, the company took on the more challenging task of assisted driving functions and now continues to work towards higher levels of autonomous driving.

The company’s new Drive Thor platform is powered by the latest Ada Lovelace generation of GPU technology, the newest Arm-based Grace CPU technology that the company announced earlier this year, and the AI capabilities from the company’s Hopper multi-instance GPU tech for AI acceleration.

With Drive Thor, automakers can consolidate intelligent vehicle functions on a single SoC.

Collectively, these enable a whopping 2 teraflops of performance. In real world terms, that means that Drive Thor can be used to enable most of the key technology-driven functions found in a modern car — including ADAS and some autonomous driving functions, automatic parking, infotainment control, driver and passenger monitoring, and instrument cluster control — from a single platform.

This is an important advance because previously many of these functions were run by separate chips. In fact, many car makers or Tier 1 automotive suppliers would often pick chips from different vendors to run these different functions. That’s why it became relatively common to hear multiple chip vendors talk about how they won business from the same vendor and sometimes even the same models. They were being used for different functions.

With Drive Thor, Nvidia is offering car manufacturers, starting with their 2025 models, the ability to start integrating more of these functions into a single platform. Given the enormous complexity of today’s cars, that’s a potentially big win for everyone involved. For Nvidia, of course, it gains more of the business and establishes itself as an even more critical supplier. For carmakers, the integration of multiple functions into a single digital platform can help simplify the overall architecture of the car, which should lead to more reliable operation, and reduce the costs and complexity associated with combining multiple platforms into a single vehicle.

Drive Thor enables this multi-function support by running multiple different applications simultaneously through a virtualized software environment (conceptually similar to what servers do in data centers and in the cloud) that Nvidia refers to as multi-domain computing.

In essence, this allows Drive Thor to replicate in software the many physical domain controllers that have powered various automotive functions within cars for decades. Also, unique to Drive Thor versus previous generation Nvidia automotive hardware is an inference transformer engine, which the company claims can deliver a 9x improvement running the types of neural networks that are at the heart of assisted and autonomous driving functions. Rounding out the news, Nvidia also announced that China car company Zeeker (part of the Geely Automotive Group, which also owns Volvo) will be the first to use Drive Thor as its computing platform starting with cars built in 2025.

At its Automotive Investor Day, Qualcomm also announced (though did not detail) a new automotive technology offering called the Snapdragon Ride Flex SoC. Like Nvidia’s Drive Thor, Qualcomm’s Ride Flex is supposed to be powerful enough to support all the key compute-powered driving functions in a smart, connected car, but full details on the chip aren’t expected until CES 2023 on January.

Qualcomm has also been releasing products for the automotive industry for many years now and, back in early 2022, launched a renewed automotive strategy with the Snapdragon Digital Chassis program. Digital Chassis integrates multiple different functions, such as Snapdragon Digital Cockpit for in-vehicle infotainment and Snapdragon Ride for assisted and autonomous driving, into a single platform that carmakers can either pick and choose from, or take as one — a path that may become more popular with the debut of the Ride Flex SoC.

Qualcomm’s history with the car industry goes much further back, as it’s been delivering communications peripherals such as cellular modems, Wi-Fi, and Bluetooth chips for cars for several decades. Most telematics systems like GM’s OnStar and its equivalent, are typically powered by Qualcomm’s connectivity chips. Plus, given the increasingly software-defined nature of modern cars, as well as the critical need and potential new business opportunity for over-the-air upgrades, the “connected” part of the smart, connected car is becoming significantly more important.

As a result of those long-term connections, the company announced at its investor event that its pipeline for automotive revenue had grown from $19 billion to $30 billion over the course of just a few months, highlighting the momentum it is seeing with partners like Acura, Audi, BMW, Cadillac, Jaguar, and Stellantis. In addition, Qualcomm unveiled a partnership with Mercedes Benz to bring the Digital Cockpit functions and the Snapdragon Automotive Connectivity Platform into future Mercedes vehicles.

For its part, Intel announced plans to spin out its Mobileye division in a widely anticipated IPO. Mobileye has been the early leader in assisted driving functions through camera-driven and smart map-based technology and has established partnerships with virtually all the large car manufacturers around the world. Recently, the company added support for lidar and other sensors with its True Redundancy platform and is increasingly moving towards more autonomous driving functions.

When Intel purchased Mobileye just over 5 years ago, it was one of the first signs of the growing importance of the automotive sector for big semiconductor companies. Since then, Intel has continued to nurture the company and helped it to grow. Now it hopes to be able to “cash in” on its investment with the newly announced IPO and leverage some of those funds for its own aggressive plans for new chip manufacturing sites.

Smart, connected cars are going to be an incredibly important part of the future growth of not only the big chip companies like Nvidia, Qualcomm, Intel, and AMD, but of lesser-known ones

As with many long-term predictions, not everything about how the big semiconductor companies and the automotive industry would work together has panned out as initially expected. After all, we are still a very long way from fully autonomous Level 5 cars for individuals — despite the tech industry’s strong prognostications and huge efforts.

However, there’s no question that smart, connected cars are going to be an incredibly important part of the future growth of not only the big chip companies like Nvidia, Qualcomm, Intel, and AMD, but of lesser-known ones like Marvell, Lattice Semiconductor, and many others.

In fact, recent reports from large investment banks have suggested that 45% of a car’s components could be tech-related in the next 10 years or so. As the high tech link to the automotive industry continues to evolve, we’re also starting to witness huge growth in cloud connected services for cars. In sum, it’s a great (and telling) example of the kind of influence that semiconductors and tech can have on traditional industries overall. It also may be a foreshadowing of other interesting re-invention opportunities still to come.

Leave a Comment

Your email address will not be published. Required fields are marked *