Qualcomm can earn money from a phone you buy without putting its name on the case. It can sell the manufacturer a Snapdragon processor or a modem. It can also license the wireless inventions that help the device communicate. One transaction supplies hardware. The other grants rights to intellectual property. That distinction explains far more about this company than another photograph of a shiny chip.
It also explains the stakes of Qualcomm’s latest expansion. The San Diego company wants the computing inside cars, laptops and industrial equipment to become a much bigger business. Its pitch draws on years spent making phones handle more work without exhausting their batteries. The question is whether that experience travels as profitably as the company hopes.
The starting position is substantial: $44.284 billion in revenue in fiscal 2025. Yet phones remain the mainstay of its semiconductor operation. Qualcomm’s story is about an incumbent trying to widen its market while some of its customers learn to build more of the technology themselves.
Two ways into the same phone
Qualcomm’s chip business, QCT, sells the processors, modems and related systems that equipment makers incorporate into finished products. Its licensing business, QTL, lets manufacturers use patented technology. The two activities are related, but their revenues are reported separately. A manufacturer’s choice of chip supplier and its need for patent rights are distinct commercial questions.
The licensing operation is unusually consequential. Qualcomm says it has more than 200 5G licensing agreements, including multiyear agreements with every major handset manufacturer. Its portfolio reaches beyond cellular radio into areas including imaging, location, audio and computing. An invention can be licensed across many manufacturers’ products; a particular chip still has to win its place in each design.
Samsung illustrates the arrangement. In 2022, the companies announced a patent agreement extending through 2030, covering 3G, 4G, 5G and forthcoming 6G technology. They also expanded their Snapdragon collaboration. Samsung can buy Qualcomm technology while developing its own processors. In this industry, customer, partner and competitor can describe the same company.
Research and development expense. About one dollar in five of annual revenue.
A roomful of engineers, then a world of calls
In 1985, seven former Linkabit colleagues gathered at Irwin Jacobs’s home to start a company. The founders were Jacobs, Andrew Viterbi, Franklin Antonio, Adelia Coffman, Andrew Cohen, Klein Gilhousen and Harvey White. The name compressed “Quality Communications” into something that would fit on a business card.
Jacobs later recalled imagining that the company might eventually employ 100 people. “Clearly, I was wrong,” he said in a 2020 company account. It is a pleasingly dry line from someone whose ambitions for the technology proved considerably less modest than his staffing forecast.
One early commercial application was communications for trucking fleets. The cellular breakthrough was CDMA, or code division multiple access. It addressed a stubborn problem: how to let more people communicate over limited radio spectrum. Qualcomm demonstrated an early CDMA call in 1989; commercial deployment followed in 1995.
Those dates are the useful part of the origin story. A technical demonstration and a commercial market were separated by years. Wireless inventions need equipment, standards, operators and manufacturers to move together. Qualcomm’s expertise grew around making a whole communications system function, rather than treating the processor as an isolated object.
“Clearly, I was wrong.”
Irwin Jacobs, recalling his early expectation of reaching 100 employees, 2020
Snapdragon packs the suitcase
Snapdragon, introduced in 2007, gave Qualcomm’s computing platforms a consumer-facing identity. The modern portfolio combines several kinds of work: general computing, graphics, AI processing and, depending on the platform, cellular connectivity. Camera processing and wireless communication have to coexist with games, video calls and the unglamorous requirement that a device stay usable away from a charger.
For phone makers such as Samsung, Xiaomi and OnePlus, buying a platform means obtaining a set of capabilities around which to design a product. For buyers, those capabilities appear as photography, responsiveness, connectivity and battery behavior. The badge alone cannot tell you how good the finished phone will be; the manufacturer’s cooling, software and component choices still matter.
Qualcomm’s advantage is this combination of computing and communications knowledge, supported by software and a large patent portfolio. MediaTek offers competing mobile platforms. Samsung and Apple develop internal silicon. In PCs, Intel and AMD bring established products and software compatibility. There is no single contest called “chips” that one company wins everywhere.
Phones still carry the weight.
The $1.4 billion ticket into custom CPUs
Qualcomm’s PC ambitions required more than putting a phone processor into a larger enclosure. In 2021, it announced an agreement to acquire CPU designer NUVIA for approximately $1.4 billion before adjustments. The acquisition brought a team focused on high-performance processor design and power efficiency.
That work helped underpin Oryon, Qualcomm’s custom CPU technology, and the Snapdragon X Elite PC platform announced in 2023. Commercial Windows laptops followed in 2024. Microsoft’s Snapdragon-powered Surface Copilot+ devices gave the initiative a visible home inside the Windows ecosystem.
The commercial logic is straightforward: laptop buyers also care about the relationship between performance and power. But moving into PCs adds another test. An organization evaluating a new machine should run its actual applications, peripherals and management tools on it. A processor benchmark is a useful measurement; it is an incomplete purchasing decision.
NUVIA’s announced price is a concrete cost of changing direction. It bought technical capability for a broader computing strategy. It did not buy guaranteed PC demand. Qualcomm must still persuade manufacturers to ship attractive systems and users to choose them.
The car is a longer conversation
Automotive gives Qualcomm another place to sell connectivity and computing together. Its Snapdragon Digital Chassis portfolio covers connected-car functions, digital cockpits and driver assistance. The opportunity extends beyond the dashboard screen: cameras and other sensors generate information that onboard computers must interpret.
In September 2025, Qualcomm and BMW unveiled Snapdragon Ride Pilot in the new BMW iX3. Their software collaboration involved more than 1,400 specialists over three years. Qualcomm contributed perception technology, while the driving policy engine was co-developed with BMW. The resulting system was also offered to other automakers and Tier-1 suppliers.
That is a different sales cycle from launching another handset chip. A driving system has to work within an automaker’s engineering, validation and production program. Qualcomm’s Ride lineup supports different levels of assistance; the presence of its silicon does not make every vehicle autonomous.
The BMW example offers a practical lesson for other suppliers: develop something specific with a demanding customer, then make the resulting platform available more broadly. The three-year effort also shows the price of admission. Reuse becomes valuable after the hard integration work, not before it.
The television that went dark
Qualcomm has already learned that technical capability does not secure every adjacent market. FLO TV tried to build a mobile television business around a dedicated broadcast network. The service and network shut down on March 27, 2011. Qualcomm later completed the sale of substantially all its associated 700 MHz spectrum to AT&T for about $1.9 billion.
The accounting is sobering. Its later proxy statement reports just $5 million in FLO TV revenue in fiscal 2011 and a $507 million loss from discontinued operations before income taxes. Those figures describe that fiscal year, not the venture’s complete lifetime cost. The spectrum sale subsequently produced a substantial gain.
Here, the service business failed to endure, while the underlying spectrum retained value. Qualcomm’s decision was observable: close the network and sell the airwaves. There is no need to invent a private moment of executive revelation. The shutdown and disposal tell the story.
For readers building technology businesses, FLO TV is the corrective to a seductive assumption. Owning useful infrastructure does not establish that customers want the service wrapped around it. A reusable asset and a viable consumer proposition require different evidence.
Dragonwing leaves the phone aisle
Qualcomm introduced Dragonwing in February 2025 to distinguish its industrial and embedded IoT, networking and cellular infrastructure offerings. The brand gives factory equipment, cameras, robots and network devices a clearer identity alongside consumer-facing Snapdragon. The shared ingredients remain AI, computing and connectivity.
On September 1, 2026, Qualcomm announced Dragonwing Q-2390 and IQ-2390 processors. The first targets connected commercial and consumer equipment; the second emphasizes industrial applications such as machine vision and automation. Evaluation kits were expected in early 2027. An announcement and a customer-ready development kit belong at different points on the calendar.
The latest financial update suggests diversification is moving. For fiscal Q3 2026, Qualcomm reported $9.9 billion in total revenue and 28% combined year-over-year growth in QCT automotive and IoT revenue. Management also described a $40 billion non-handset revenue target for fiscal 2029. That is an ambition, not money already collected.
Start with the workload
Developers can engage with Qualcomm well before negotiating a chip purchase. AI Hub provides models, sample applications and tools for optimizing and testing on-device AI. Its Workbench can profile models on hosted Qualcomm devices, helping teams examine latency and memory requirements before committing to hardware.
A sensible starting point is one useful task: recognizing an object, processing audio or running a small assistant. Choose a model, measure it on the intended device and check the result against the product’s constraints. Local processing can reduce reliance on a network connection, but the model still has to fit the available memory and power budget.
Qualcomm’s open-source presence adds another entry point. Its GitHub organization directs developers to AI tools, Linux work, audio software and robotics projects. This is part of the practical value of a platform supplier: customers need working code and development tools as well as component specifications.
The broader business lesson is to identify a recurring constraint and reuse the expertise that solves it. Qualcomm has done that with wireless capacity and power-efficient computing. The limits remain visible: customers can design competing chips, software can complicate adoption, and an adjacent market can reject a service. Its next chapter will be decided inside products people choose to use.