Qualcomm's Third Gamble

In the third quarter of fiscal 2026 (April to June), Qualcomm's automotive business revenue reached 1.588 billion US dollars, a year-on-year increase of 61%, maintaining double-digit growth for 23 consecutive quarters.

Automotive chips, once regarded as a "side business", have grown into the second growth curve of this mobile chip giant. However, to support this huge ambition, simply selling more cockpit chips is far from enough. The next battlefield Qualcomm is betting on is the integrated cockpit-driving platform.

At the end of 2025, BMW's new-generation iX3 went into mass production. Hidden in this vehicle is also Qualcomm's intelligent driving system Snapdragon Ride. To develop this system well, Qualcomm and BMW spent three years and invested more than 1,400 professionals.

For Qualcomm, intelligent driving is an unfamiliar business. More than a decade ago, it could package chips, software and design together, enabling a mobile phone manufacturer to launch products in as fast as 60 days; later, it moved this product capability into the automotive industry, and captured the intelligent cockpit market with the 8155 chip.

From mobile phones to cockpits, Qualcomm has won twice. What it wants to do now is to integrate the cockpit and intelligent driving into the same computing platform, and win for the third time.

But this time, replicating past successful experiences to the new AI-driven battlefield is far more difficult than ever.

A demonstration that almost failed

In November 1989, Qualcomm held the first public demonstration of CDMA mobile communication in San Diego.

Outside the venue, a test car was prepared to drive between two base stations to prove that calls could be switched continuously. Before the demonstration started, the system suddenly malfunctioned. On the stage, 56-year-old founder Irwin Jacobs saw the engineers gesturing constantly, so he could only temporarily extend his speech by 45 minutes to win time for the team. The equipment finally recovered, and the test car successfully completed the call.

Qualcomm later called it "the moment that made Qualcomm". Jacobs' recollection is even more thrilling: "If that demonstration failed, we might be finished."

At that time, Qualcomm had been established for only four years, and the United States was about to enter the 2G era. The mainstream opinion preferred TDMA, which was easy to understand and verify, while Qualcomm bet on the more complex CDMA. To convince the market, Qualcomm not only had to propose communication algorithms, but also developed chips, mobile phones, base stations and system software together to make CDMA work truly.

Irwin Jacobs' efforts eventually led the United States to accept two sets of 2G standards. For Qualcomm, this experience shaped its most important capability later: to find a bottom-level technology that customers cannot bypass, and then turn it into a system that can be commercialized on a large scale.

In the smartphone era, this fulcrum became the baseband.

The baseband is responsible for enabling mobile phones to access operator networks in accordance with communication standards. It needs to support different systems and global frequency bands, as well as handle switching, signals and power consumption. Only a few companies in the world have complete capabilities, and Qualcomm is one of them.

Because Qualcomm happened to stand at both ends of standards and products at the same time, deeply participated in CDMA, 3G and LTE, and developed global multi-mode basebands earlier. Later, when facing antitrust investigations, market data cited by the U.S. Court of Appeals for the Ninth Circuit showed that from 2006 to 2016, Qualcomm's share in the CDMA baseband market exceeded 90%; from 2011 to 2016, it accounted for at least 70% of the high-end LTE baseband market.

But the baseband advantage will not automatically turn into mobile phone chip market share. From 2012 to 2014, Qualcomm's annual report listed more than ten direct competitors, including Texas Instruments, NVIDIA, Broadcom and Intel.

Qualcomm's solution is to integrate baseband, CPU, GPU, positioning, audio and video and other functions into the Snapdragon platform, while providing operating system adaptation, driver and operator certification support. Mobile phone manufacturers no longer need to purchase scattered components for integration by themselves, but directly buy a complete set of connected global solutions, namely SoC.

This set of strategies is not easy to replicate. Texas Instruments and Broadcom failed to cross the baseband threshold; NVIDIA's Tegra had GPU advantages but lacked mature basebands. After acquiring the British baseband chip company Icera in 2011, it still failed to build a complete mobile platform, and closed its baseband business in 2015. MediaTek has long been better at the mid-to-low end, and its high-end baseband, imaging and flagship customer follow-up are relatively slow. The high-end chips of Apple, Samsung and Huawei are mainly for self-use, so Qualcomm has become one of the few flagship platforms that Android manufacturers can purchase publicly.

For small and medium-sized manufacturers, Qualcomm also provides QRD public version design. In 2013, Qualcomm stated that some customers could launch a mobile phone in as fast as 60 days. For mobile phone startups in those years, what Qualcomm sold was certainty in a rising industry, a mobile phone that could be placed on the counter two months later.

Qualcomm has proved in the mobile phone industry that chip competition is not only about performance. Whoever can solve the most complex and most delay-prone parts in advance can reduce the mass production risk of customers.

Later, Qualcomm brought this method into automotive cockpits.

Capturing the automotive cockpit market

In 2022, Zeekr did something uncommon in the industry.

In July of this year, Zeekr announced that it would provide a free upgrade to Qualcomm 8155 for all delivered and ordered Zeekr 001 vehicles. Previously, the infotainment system lag and slow startup of the first batch of vehicles continuously caused dissatisfaction among car owners. Zeekr finally replaced the complete cockpit system. From hardware procurement, software re-development to logistics and after-sales, the total investment exceeded 300 million yuan.

Zeekr was willing to spend 300 million yuan because the infotainment system experience was enough to affect the overall reputation of a smart electric vehicle, and 8155 was one of the high-end cockpit chips that had been verified by large-scale mass production and had low risks at that time.

By 2023, Qualcomm's share in China's cockpit domain control chip market reached 59.2%; after 8295 took over in 2024, the share further rose to 70%. A number of brands including Geely, Great Wall, NIO, XPeng, Li Auto, Geely, Cadillac and Smart have all used 8155 chips.

Qualcomm did not suddenly break into the automotive industry with mobile phone chips. As early as 2002, it relied on communication technology to provide CDMA in-vehicle solutions for General Motors; by the time it released the first generation of cockpit chips in 2014, Qualcomm's communication solutions had entered about 10 million vehicles. The communication business first proved its automotive-grade quality, long-term supply capacity and technical strength for it.

However, Qualcomm's first two generations of cockpit chips did not sell much, and the traditional manufacturers such as Renesas and NXP were still the market leaders. 602A and 820A were more like admission tickets: helping Qualcomm complete automotive-grade and mass production verification, and gradually allowing Tier 1 suppliers such as Bosch, Continental and Desay SV to get familiar with the Snapdragon platform.

The turning point came in 2019. Qualcomm released the third-generation cockpit platform represented by 8155, which just caught up with the concentrated upgrade of smart cockpits in China's new energy vehicles.

In the era of fuel vehicles, the cockpit is mainly responsible for navigation, audio and video, and vehicle settings. For new energy vehicles, large screens, multi-screens, 3D interaction, online applications, voice assistants and OTA have gradually become selling points. The cockpit has also evolved from an infotainment system to the most important human-computer interaction entrance of the whole vehicle. As a result, automakers need high-performance SoCs to handle system operation, graphics rendering, audio and video, connectivity, cameras and end-side AI at the same time.

These are exactly what Qualcomm has been doing in the mobile phone industry for more than ten years.

At that time, not only Qualcomm was making cockpit chips. Samsung had Exynos Auto V9, MediaTek had MT8666, and Renesas also had R-Car. The 8155 did not maintain a leading position in every parameter, but in the key window from 2019 to 2021, the 8155 was the only product that met performance, power consumption, software adaptation and mass production requirements at the same time without obvious shortcomings.

Samsung entered the market and achieved mass production later, mainly serving European automakers; MediaTek's MT8666 at that time was more mid-range; Renesas, which knew the automotive industry best, had a significantly slower product iteration pace than consumer electronics. Qualcomm just caught the demand of China's new energy vehicles for flagship cockpits.

More importantly, Qualcomm brought the platform-oriented strategy from the mobile phone era to the automotive industry. The third-generation cockpit platform covers different grades, and automakers can replace chips in models with different prices ranging from 100,000 to 300,000 yuan, while reusing the operating system, voice, navigation, middleware, application store and OTA system, without redeveloping a complete set of cockpit for each model. This is also the advantage that Samsung, MediaTek and other players did not have in 2019.

After the first two generations of products paved the way, more and more suppliers have completed mass production around Qualcomm's platform. After automakers choose 8155, they can choose from Tier 1 suppliers such as Bosch, Continental, Desay SV, LG and Visteon. When compatibility and engineering problems occur, it is also easier to find teams with similar experience, which further reduces the concerns of automakers.

As a result, 8155 has developed from a chip deep in the supply chain to an industry default platform that automakers frequently write into configuration tables and consumers can directly name the model of.

Expensive catch-up lessons

In 2021, the year when 8155 achieved large-scale volume release, Qualcomm took a fancy to another automotive supply chain company.

In July, Magna had reached an acquisition agreement with Swedish automotive safety supplier Veoneer. Less than a month later, Qualcomm offered 37 US dollars per share, 18% higher than Magna's offer, and publicly joined the bidding. In the end, Qualcomm joined hands with SSW Partners to acquire Veoneer, and took Arriver, which it really wanted, into its pocket.

This time Qualcomm wanted to enter the intelligent driving field. Qualcomm can reuse a lot of technology from the mobile phone era in the cockpit, but it has to start from scratch when entering the intelligent driving field. Arriver, which focuses on computer vision, perception and driving strategy, can make up for this deficiency very well.

The direct reason for Qualcomm to enter the intelligent driving market is that intelligent driving can bring higher revenue.

In 2022, Qualcomm predicted that the addressable automotive market would be about 100 billion US dollars by 2030, of which intelligent driving would account for about 59 billion US dollars and cockpits about 25 billion US dollars; for high-end models, Qualcomm hopes to package and sell cockpits, intelligent driving, connectivity and software, so that the value of per vehicle content can be increased to about 3,000 US dollars at most.

In addition, if Qualcomm does not develop intelligent driving and focuses on the cockpit market, it may not be able to hold on. Because the automotive computing architecture is changing. In the past, cockpit and intelligent driving chips were separated. Qualcomm sold 8155 and NVIDIA sold Orin, and the two could appear in the same vehicle at the same time. If the market moves towards integrated cockpit-driving in the future, the cockpit and intelligent driving may share one computing platform. NVIDIA Thor is designed to carry intelligent driving, parking, instrument and infotainment at the same time from the very beginning, and Horizon, Huawei and other players are also developing similar solutions. If Qualcomm only sticks to the cockpit market, the sand in its hand will only get less and less.

Therefore, the integrated cockpit-driving solution is not only an offensive move for Qualcomm, but also a defensive move.

At present, in the 100,000-200,000 yuan market, Qualcomm's integrated cockpit-driving solution has been mass-produced. Some versions of Arcfox Alpha T5, S5 and Dongfeng Nissan N6 have used one Qualcomm 8775 to handle both cockpit and intelligent driving; the higher-priced Leapmotor D19 uses two 8797 chips to integrate the cockpit and intelligent driving into the same architecture.

But the competition of this new technology architecture has just begun. Qualcomm's self-developed intelligent driving software Snapdragon Ride Pilot currently has only one mass-produced and delivered model, the BMW iX3. There is still a long way to go to form the same scale and reusability as the 8155 chip.

In the new field, the weapons Qualcomm accumulated in the past are not handy. Qualcomm has almost dealt with Android, graphics, audio and video and connectivity required by cockpits in the mobile phone era; high-level intelligent driving needs to continuously process camera and radar data, run complex AI models, and meet real-time performance and functional safety requirements at the same time. These are new capabilities to be conquered.

More importantly, NVIDIA has formed a strong AI development ecosystem around CUDA. Although Qualcomm has AI Stack and Ride toolchain, it has not formed similar industry inertia.

Therefore, in the Chinese market, Qualcomm's intelligent driving revenue mainly comes from hardware. Arcfox uses Qualcomm chips with Zongyu, Dongfeng Nissan N6 is equipped with Momenta, and the software of Leapmotor D19 is developed by Leapmotor itself. Openness is Qualcomm's advantage, and automakers do not need to hand over the algorithms to chip companies. Conversely, it also means that Qualcomm has not yet built a set of development systems that customers can hardly bypass, as it did in the mobile phone and cockpit eras.

Intelligent driving software requires large investment, updates quickly, and faces fierce market competition. It is not easy for Qualcomm to form a moat in the intelligent driving market. NVIDIA, Mobileye, Horizon, Huawei and even the self-research teams of automakers are fighting fiercely in this battlefield.

More than 30 years ago, Qualcomm found a fulcrum that the industry could not bypass through the underlying mathematical model of CDMA; in the 8155 era, it occupied the cockpit market with mature SoC engineering and supply chain agility. In the intelligent driving arena, huge computing power, massive data, and continuously evolving model capabilities are the core competitiveness. This brand new game triggered by new technologies has just begun.

This article is from the WeChat official account "Yunjian Insight", written by Liu Yimo, and published with authorization from 36Kr.