Qualcomm

Qualcomm Incorporated
Type	Public
Traded as	Nasdaq: QCOM; S&P 500 component
Industry	Semiconductors; Telecommunications; Wireless technology licensing
Founded	July 1985, San Diego, California
Founder	Irwin Jacobs, Andrew Viterbi, Harvey White, Franklin Antonio, Adelia Coffman, Andrew Cohen, Klein Gilhousen
Headquarters	San Diego, California, United States
Key people	Cristiano Amon (President and chief executive officer)
Products	Snapdragon systems on chips, modems, radio frequency front ends, automotive platforms, patent licensing
Employees	Approximately 49,000
Website	qualcomm.com

Qualcomm Incorporated is an American semiconductor and telecommunications company headquartered in San Diego, California. It designs the processors and modems used in a very large proportion of the world's smartphones, and it owns the patents underlying the cellular standards that every smartphone must implement.

The company was founded in July 1985 by seven people, led by Irwin Jacobs and Andrew Viterbi, both of whom had previously founded Linkabit and both of whom were among the most significant figures in the theory of digital communication. Its name is a contraction of Quality Communications.

Qualcomm's business has two halves whose relationship is the central fact about the company. QCT designs and sells chips, principally the Snapdragon family, and produces most of the revenue. QTL licenses the company's patents to essentially every manufacturer of cellular equipment and produces most of the profit, because the royalties are collected on the value of the device rather than on the value of the chip and because the underlying patents are essential to standards that cannot be implemented without them.

That arrangement has made Qualcomm exceptionally profitable and has made it the subject of antitrust investigation and litigation on four continents, including proceedings in China, South Korea, the European Union, Taiwan and the United States, and a two-year dispute with Apple that was settled on the morning the trial began.

The company has diversified substantially since 2021, into automotive computing, where it is among the fastest-growing suppliers, into personal computers with the Snapdragon X series, into connected devices, and most recently into data centre and artificial intelligence infrastructure through the acquisition of Alphawave and others.

History

CDMA, 1985 to 1999

Irwin Jacobs and Andrew Viterbi had founded Linkabit in 1968, a communications company that produced a considerable number of subsequent San Diego technology firms. After selling it to M/A-COM they founded Qualcomm in July 1985 with five colleagues, meeting initially in Jacobs's living room.

The company's early work was in satellite communications, producing OmniTRACS, a messaging and position-reporting system for long-haul trucking that was commercially successful and that funded what followed.

The decisive commitment was to code division multiple access, a method of sharing radio spectrum that differs fundamentally from the alternatives. In time division, each user is allocated a slot; in frequency division, a channel. In CDMA every user transmits across the whole band simultaneously, each encoded with a distinct spreading sequence, and the receiver recovers the wanted signal by correlating against its sequence while the others appear as noise.

The theoretical advantage is capacity, because the system degrades gracefully rather than blocking when demand rises, and because it exploits the statistical fact that people do not talk continuously. The practical objection, advanced by essentially the entire industry, was that it required power control accurate to a fraction of a decibel updated hundreds of times a second, because a transmitter close to the base station would otherwise drown those further away. Engineers at the established equipment makers argued that this could not be done. The reaction is remembered in the industry through the remark attributed to several sources that CDMA violated the laws of physics.

Qualcomm demonstrated it working in San Diego in November 1989. CDMA was adopted as the IS-95 standard in 1993 and deployed commercially from 1995, and it became the basis of one of the two families of third-generation systems, and, through WCDMA, of the other as well, which meant that Qualcomm held essential patents whichever path the industry took.

The company also built handsets and infrastructure in the 1990s, selling the base station business to Ericsson in 1999 as part of a settlement of patent litigation, and the handset business to Kyocera in the same year. Both disposals were strategic: Qualcomm concluded that it should sell components and licences rather than compete with its own customers.

Snapdragon and the smartphone, 2000 to 2015

Qualcomm's third-generation and then fourth-generation modems placed it at the centre of the smartphone as the category emerged.

The Snapdragon brand, introduced in 2007, covers systems on a chip combining an application processor, graphics, a modem, signal processing and connectivity. The company acquired the graphics assets that became Adreno from AMD in 2009, which is why the name is an anagram of Radeon, and developed its own processor cores under the Krait and Kryo names in addition to licensing Arm designs.

Its position through the 2010s was very strong and at times close to exclusive. Android manufacturers used Snapdragon almost universally at the high end, and Apple, which designed its own application processors, bought Qualcomm modems because nothing else worked as well.

The licensing business grew alongside. Qualcomm's royalty is calculated as a percentage of the wholesale price of the complete device, subject to a cap, rather than of the chip. The company's justification is that its patents cover the system-level invention that makes the device work as a phone, not merely a component. Manufacturers' objection, which produced most of the litigation described below, is that this collects a share of the value of cameras, displays and storage that Qualcomm did not invent.

Broadcom, NXP and the antitrust years, 2015 to 2020

The period from 2015 was the most difficult in the company's history and it faced three simultaneous problems.

Regulators. China's National Development and Reform Commission fined Qualcomm approximately US$975 million in February 2015 and required changes to its Chinese licensing, including a reduced royalty base. South Korea's Fair Trade Commission fined it approximately US$854 million in December 2016 for refusing to licence competing chipmakers and for conditioning chip supply on patent agreements. The European Commission fined it €997 million in January 2018 for payments to Apple in exchange for exclusivity, and €242 million in July 2018 for predatory pricing. Taiwan's authority imposed and then settled a very large fine.

The United States Federal Trade Commission sued in 2017, and Judge Lucy Koh ruled against Qualcomm in May 2019, finding that its practice of refusing to sell chips to manufacturers who had not taken a patent licence, which the industry calls no licence, no chips, was anticompetitive. The Ninth Circuit reversed comprehensively in August 2020, holding that Qualcomm's conduct was hypercompetitive rather than anticompetitive and that the district court had erred in treating a refusal to license competitors as an antitrust violation. The reversal was the most important legal event in the company's history.

Apple. Apple sued in January 2017 alleging that the royalties were excessive and that Qualcomm had withheld rebates in retaliation for Apple's cooperation with regulators. Qualcomm counter-sued, sought import bans, and the dispute escalated across several jurisdictions and involved Apple's contract manufacturers withholding payments. Apple moved its modem supply to Intel. The two settled on 16 April 2019, the morning opening statements were due, with Apple paying a reported sum, taking a six-year licence and a multi-year chip supply agreement. Intel announced its exit from smartphone modems the same day, having concluded that it could not compete, and Apple bought Intel's modem business in July 2019 for approximately US$1 billion in order to build its own.

Broadcom. Broadcom made a hostile approach in November 2017, raised to approximately US$121 billion, which would have been the largest technology acquisition ever attempted. Qualcomm rejected it, and the Committee on Foreign Investment in the United States intervened on national security grounds, with President Donald Trump prohibiting the transaction by executive order in March 2018. The stated concern was that Broadcom, then domiciled in Singapore, would cut research spending and thereby cede leadership in 5G standards to Huawei.

Qualcomm's own attempt to acquire NXP Semiconductors for approximately US$44 billion, agreed in 2016, was abandoned in July 2018 after Chinese regulatory approval was not granted, in a period of escalating trade conflict. Qualcomm paid a US$2 billion break fee. The failure was strategically significant, because NXP would have given the company the automotive and industrial position it subsequently had to build itself.

Diversification, 2021 to 2026

Cristiano Amon, a Brazilian engineer who had joined Qualcomm in 1995 and run the chip business, became chief executive in June 2021. His stated strategy has been to reduce the company's dependence on smartphones, which had made it exceptionally profitable and exceptionally exposed.

The exposure was concrete. Apple was developing its own modem and would eventually stop buying, Samsung alternates between Snapdragon and its own Exynos, and the Chinese manufacturers who represent a large share of Android volume are subject to trade restrictions and are developing domestic alternatives. Qualcomm has disclosed that it expects its Apple business to decline substantially, and Apple's first internally designed modem shipped in 2025.

Nuvia, a chip design company founded by former Apple processor architects including Gerard Williams III, was acquired in January 2021 for approximately US$1.4 billion. Its work produced the Oryon processor cores, which underpin the Snapdragon X series for personal computers, introduced in 2024 and representing Qualcomm's most serious attempt to enter a market dominated by Intel and AMD. The Arm-based Windows machines built on them offered battery life substantially better than x86 equivalents, and their adoption has been constrained by software compatibility.

The acquisition produced a serious dispute. Arm contended that Nuvia's architecture licence did not transfer to Qualcomm and moved in October 2024 to cancel Qualcomm's design licence. A United States federal jury ruled largely in Qualcomm's favour in December 2024, finding that the designs were properly licensed under Qualcomm's own agreement, and litigation between the two has continued in modified form.

Also in October 2024, the exclusivity arrangement under which Qualcomm was the sole supplier of Arm processors for Windows expired on 25 December 2024, opening the category to NVIDIA, MediaTek and others.

Automotive has been the most successful diversification. Qualcomm's Snapdragon Digital Chassis covers infotainment, telematics, driver assistance and automated driving, and the company has disclosed a design win pipeline measured in tens of billions of dollars. It acquired Arriver, the software arm of Veoneer, in 2022, and Autotalks, an Israeli specialist in vehicle-to-everything communication, completing on 5 June 2025 after a review by the British Competition and Markets Authority and an initial abandonment in 2024 following a Federal Trade Commission inquiry. In September 2025 Qualcomm and BMW announced a jointly developed automated driving system debuting on the BMW iX3.

Data centre and artificial intelligence became a priority from 2024, when the company identified generative artificial intelligence at the edge as a principal demand driver and projected a total addressable market of approximately US$900 billion by 2030. Reuters reported in August 2024 that Qualcomm had examined parts of Intel's business. The acquisition of Alphawave IP Group, a British designer of high-speed wired connectivity, was agreed in June 2025 for US$2.4 billion specifically to enter data centre and artificial intelligence infrastructure, where the constraint is increasingly the movement of data between chips rather than the computation itself.

Further acquisitions have included Movian AI, the generative artificial intelligence unit of the Vietnamese research company VinAI, in April 2025; Arduino, the Italian maker of single-board microcontrollers, in October 2025, announced alongside the Arduino Uno Q using a Qualcomm system on a chip and intended for artificial intelligence and robotics; Ventana Micro Systems, a designer of processors based on the open RISC-V instruction set, in December 2025, which both strengthens the company's processor engineering and provides an alternative to Arm given the licensing dispute; and Modular, for US$3.9 billion in 2026.

In March 2026 Qualcomm launched a chip aimed at new categories of Android device, addressing renewed interest among technology companies in hardware form factors beyond the phone.

The company also agreed in September 2023 a sponsorship reported at approximately US$75 million a year for Snapdragon to be the principal shirt sponsor of Manchester United from the 2024-25 season, replacing TeamViewer, an unusually large consumer marketing commitment for a component supplier and one intended to establish Snapdragon as a brand consumers recognise.

Business

QCT

Qualcomm CDMA Technologies designs and sells semiconductors and represents the large majority of revenue.

Snapdragon systems on a chip range from entry-level parts to the flagship 8 series used in premium Android phones. A modern Snapdragon integrates Arm-based or Oryon processor cores, an Adreno graphics processor, a Hexagon digital signal processor used increasingly for machine learning, an image signal processor, a modem and connectivity.

Modems are sold separately as the X series, and the radio frequency front end, comprising filters, amplifiers and switches, was built into a substantial business through the 2016 RF360 joint venture with TDK, subsequently bought out, because a modem is of limited use without the analogue components that connect it to an antenna and because integrating the two improves performance.

Qualcomm is fabless and has its chips manufactured by TSMC and Samsung, alternating between them.

Adjacent markets now include automotive, personal computers, extended reality headsets, where Qualcomm supplies essentially the entire industry including Meta's Quest devices, networking, industrial and connected devices.

QTL

Qualcomm Technology Licensing licenses the company's portfolio, which comprises well over one hundred thousand patents and applications covering 3G, 4G and 5G standards and a considerable quantity of adjacent technology.

The licences are worldwide and cover essentially every manufacturer of cellular devices. Royalties are calculated on the selling price of the device, subject to a cap that has been reduced over time under regulatory pressure, and the rates are published in outline following the settlements with regulators.

QTL's revenue is a fraction of QCT's and its operating margin is far higher, because the marginal cost of a licence is zero, so the two segments contribute comparable amounts of profit. This is the arrangement that competitors and customers have spent two decades attacking and that has survived essentially intact.

The company's argument for the structure is that its research produces the standards from which the entire industry benefits, that it spends approximately US$8 billion a year on research and development, much of it on work contributed to standards bodies, and that a licensing model is the only way to be compensated for inventions that everyone must use. The objection is that standard-essential patents confer market power conferred by the standard itself rather than earned in competition, and that the holder's commitment to license on fair, reasonable and non-discriminatory terms is enforceable only through litigation that few can afford.

Technology

Cellular standards

Qualcomm's technical contribution runs through every generation of cellular technology since the second.

CDMA and the IS-95 standard established the company. The engineering problems it solved, fast closed-loop power control, the rake receiver which combines multipath reflections that other systems treat as interference, soft handover between base stations, and variable-rate vocoders that reduce transmission during silence, are each substantial in their own right.

3G was built on CDMA in both its principal forms, CDMA2000 and WCDMA, which meant that Qualcomm held essential patents whichever the operator chose and which is the origin of the licensing position it holds today.

4G LTE moved to orthogonal frequency division multiplexing, a different access method, which reduced Qualcomm's structural advantage without eliminating its patent position, since the company had contributed extensively and had acquired Flarion in 2006 precisely to hold OFDM patents.

5G introduced millimetre wave operation, using frequencies above 24 GHz which offer enormous bandwidth and propagate poorly, requiring beamforming and antenna arrays in the handset itself. Qualcomm's work on practical millimetre wave modules for phones, which many engineers had regarded as impossible within a handset's power and space budget, was among its more notable achievements, and the company has been a principal contributor to the standard through 3GPP.

6G work is under way, with the company positioning artificial intelligence integrated into the radio interface as a defining characteristic.

On-device artificial intelligence

Qualcomm's strategic argument since 2023 has been that a substantial proportion of machine learning inference will move from data centres to devices, for reasons of latency, cost, privacy and connectivity, and that the company's position in the processors inside those devices makes it the principal beneficiary.

The Hexagon processor, originally a digital signal processor for audio and modem work, has been progressively rebuilt for neural network inference, and recent Snapdragon parts run language models of several billion parameters on a phone.

The argument is coherent and its commercial realisation is uncertain. Running a model on a device is cheaper for whoever would otherwise pay for the server, which is not the person buying the phone, so the willingness to pay more for the capability is unclear. What is clearer is that the capability has become a requirement for competing at the high end, which supports Snapdragon's pricing regardless of whether consumers value it directly.

Products

Snapdragon

The Snapdragon brand, introduced in 2007, now covers several distinct families.

Snapdragon 8 series parts are the flagship mobile processors, used in premium Android devices from Samsung, Xiaomi, Oppo, Vivo, Honor, OnePlus, Motorola and others. The 8 Elite generation, introduced in 2024, was the first mobile Snapdragon built on the Oryon cores derived from the Nuvia acquisition, replacing licensed Arm designs in the flagship line.

Snapdragon 7, 6 and 4 series cover the mid-range and entry segments, which represent the majority of units and a minority of revenue, and where competition from MediaTek and from Chinese suppliers including UNISOC is most intense.

Snapdragon X is the personal computer family, launched in 2024 with the X Elite and X Plus, and extended since. The parts combine Oryon cores with Adreno graphics and a neural processing unit, and are used in Windows machines from Microsoft, Dell, HP, Lenovo, Asus, Acer and Samsung. Their principal advantage is performance per watt, which produces battery life comparable to Apple's and substantially better than x86 equivalents; their principal difficulty is that Windows software is overwhelmingly compiled for x86 and runs through emulation, which works well for most applications and poorly for some.

Snapdragon Digital Chassis covers automotive, with the Snapdragon Ride platform for driver assistance and automated driving, Snapdragon Cockpit for infotainment and instrumentation, and Snapdragon Auto Connectivity. Customers include BMW, Mercedes-Benz, General Motors, Stellantis, Honda, Hyundai, Volvo, Renault, Li Auto, Nio and Xpeng.

Snapdragon XR and AR parts supply the great majority of extended reality headsets, including Meta's Quest family, and the company has worked with Meta, Samsung and Google on the platforms underlying them.

Snapdragon Wear serves smartwatches, Snapdragon Sound covers audio, and the QCC and QCS families serve connected devices, robotics and industrial applications, the last strengthened by the Arduino acquisition in October 2025.

Modems and radio frequency

The modem is Qualcomm's oldest and most defensible product. A cellular modem must implement several generations of standard simultaneously, must operate across dozens of frequency bands whose allocation differs by country, must handle carrier aggregation across several of them at once, and must be certified with hundreds of operators. The certification and field-testing burden alone is a substantial barrier, and it is the principal reason Intel withdrew and Apple took most of a decade to produce its own.

The X series modems are sold discretely and integrated into Snapdragon parts. The radio frequency front end, comprising power amplifiers, filters, switches and antenna tuners, was built into a large business from the RF360 venture with TDK in 2016 and subsequent acquisitions, and it matters because modem and front end must be designed together to achieve the performance operators require.

The company also supplies Wi-Fi and Bluetooth, principally through the Atheros business acquired in 2011, and satellite connectivity through partnerships including the Snapdragon Satellite arrangement with Iridium, which was discontinued in 2023 after handset manufacturers declined to adopt it.

Ventures that did not work

Qualcomm's record includes several substantial failures, which are instructive because they were each defensible when begun.

BREW, the Binary Runtime Environment for Wireless, introduced in 2001, was an application platform for mobile phones that predated the iPhone by six years and that was commercially successful in the United States and Japan. It was controlled by the operators, who decided what could be published and took the majority of revenue, and it was swept away by application stores that dealt directly with developers. The episode is a clean demonstration that the better platform is the one with the lower barrier to publishing.

MediaFLO was a mobile television broadcasting system, for which Qualcomm bought spectrum and built a national network in the United States from 2007. It required a separate receiver, offered a fixed schedule of channels, and was launched as streaming over cellular data was becoming practical. Qualcomm closed it in 2011 and sold the spectrum to AT&T for approximately US$1.9 billion, recovering most of the investment.

Mirasol was a reflective display technology using interferometric modulation, which produced colour images readable in sunlight and consuming power only when the image changed. It was technically remarkable and could not be manufactured at adequate quality and cost, and Qualcomm wrote off a purpose-built factory in Taiwan.

Halo was wireless electric vehicle charging by magnetic resonance, developed from 2011 and sold to WiTricity in 2019.

Qualcomm Centriq, a server processor based on Arm announced in 2017, was cancelled in 2018 after Qualcomm came under pressure from Broadcom's approach and from activist investors to reduce costs, at a moment when Arm server processors were widely regarded as unviable. Amazon's Graviton, launched the same year, subsequently demonstrated that they were not, and the cancellation is generally regarded as Qualcomm's most costly strategic error. The company's 2025 and 2026 moves into data centre infrastructure are a return to the same market seven years later.

Competition

MediaTek of Taiwan is Qualcomm's principal competitor in mobile processors and has taken the larger share of global smartphone shipments by unit, competing on price in the mid-range and, with the Dimensity family, increasingly at the high end.

Apple designs its own application processors and, since 2025, its own modems, and is therefore both a large customer becoming a former customer and the benchmark against which Snapdragon is judged.

Samsung designs Exynos processors and alternates between them and Snapdragon in its own phones, a relationship that is simultaneously competitive, customer and, since Samsung Foundry manufactures some Snapdragon parts, supplier.

Chinese suppliers including UNISOC, and Huawei's HiSilicon where it can obtain manufacturing, are growing, assisted by policy directed at reducing dependence on American technology.

Intel and AMD are the incumbents Qualcomm attacks in personal computers, and NVIDIA entered the same market after the Windows exclusivity expired at the end of 2024.

Broadcom, Qorvo and Skyworks compete in radio frequency components, the last two principally in Apple's supply chain.

NVIDIA, Mobileye, Texas Instruments and NXP compete in automotive computing.

Research and standards

Qualcomm spends approximately US$8 billion a year on research and development, among the highest proportions of revenue in the semiconductor industry, and a substantial part of it is directed at standards work rather than at products.

The company maintains a large presence in 3GPP, the body that defines cellular standards, and its contributions are the origin of the patent portfolio that QTL licenses. This is the mechanism by which the business model works: invent, contribute the invention to a standard so that everyone adopts it, accept the obligation to license on fair terms, and collect.

Qualcomm Research has also worked on areas that have not yet produced products, including neuromorphic computing through the Zeroth project, quantum-resistant cryptography, and radio interfaces incorporating machine learning, which the company has positioned as a defining feature of the sixth generation.

The company's engineers have received substantial recognition, and Irwin Jacobs and Andrew Viterbi are both members of the National Academy of Engineering; Viterbi's algorithm for decoding convolutional codes, published in 1967, is used in essentially every digital communication system in existence, including those of Qualcomm's competitors.

Timeline

1985 Founded in July by seven people led by Irwin Jacobs and Andrew Viterbi.
1988 OmniTRACS is launched for long-haul trucking.
1989 CDMA is demonstrated in San Diego in November.
1993 CDMA is adopted as the IS-95 standard.
1995 The first commercial CDMA networks are deployed.
1999 The infrastructure business is sold to Ericsson and handsets to Kyocera.
2001 BREW is launched.
2005 Paul Jacobs succeeds his father as chief executive.
2007 The Snapdragon brand is introduced; MediaFLO launches.
2009 The Adreno graphics assets are acquired from AMD.
2011 MediaFLO closes; Atheros is acquired.
2014 Steve Mollenkopf becomes chief executive.
2015 China's regulator fines the company approximately US$975 million.
2016 South Korea fines it approximately US$854 million; the NXP acquisition is agreed.
2017 Apple sues; the Federal Trade Commission sues; Broadcom makes a hostile approach in November.
2018 The Broadcom bid is blocked by executive order in March; the NXP acquisition is abandoned in July; the European Commission imposes two fines.
2019 The Apple dispute is settled on 16 April; Intel exits smartphone modems; Judge Koh rules against Qualcomm in May.
2020 The Ninth Circuit reverses the FTC decision comprehensively in August.
2021 Nuvia is acquired in January; Cristiano Amon becomes chief executive in June.
2023 The Manchester United sponsorship is agreed in September.
2024 The Snapdragon X series launches; the Windows exclusivity expires on 25 December; a jury rules largely for Qualcomm against Arm in December.
2025 Autotalks completes in June; Alphawave is agreed for US$2.4 billion; Arduino is acquired in October; Ventana Micro Systems in December; the BMW automated driving system is announced in September.
2026 Modular is acquired for US$3.9 billion; a new chip for emerging Android device categories is launched in March.

Corporate affairs

Cristiano Amon has been chief executive since June 2021. He joined Qualcomm as an engineer in 1995, having previously worked in Brazilian telecommunications, and led the chip business before his appointment.

His predecessors are Steve Mollenkopf, 2014 to 2021, who managed the regulatory conflicts and the Broadcom approach and who had been offered the Microsoft chief executive role in 2013; Paul Jacobs, 2009 to 2014, the son of the founder, who left the board in 2018 after proposing a bid to take the company private; and Irwin Jacobs, 1985 to 2005.

Qualcomm employs approximately 49,000 people, with San Diego as its principal centre and substantial engineering in India, notably Hyderabad and Bengaluru, in China, Taiwan, South Korea, Israel, following several acquisitions, Austria, and in the United Kingdom through Alphawave.

The Jacobs family has been central to San Diego's civic life, and Irwin and Joan Jacobs have given very large sums to the University of California San Diego, to Cornell, to the San Diego Symphony and to the Salk Institute, among others.

Revenue has run in the region of US$35 billion to US$45 billion in recent years, with handsets the largest component, automotive and connected devices growing fastest, and licensing contributing a disproportionate share of profit. The company pays a dividend and has repurchased shares substantially.

Controversies

Beyond the regulatory proceedings described above, several matters recur.

The royalty base. Charging a percentage of the device price rather than of the chip price is the practice at the centre of every dispute. Qualcomm has reduced the cap and adjusted rates under regulatory pressure in China, South Korea and elsewhere, and has retained the structure.

No licence, no chips. The practice of declining to supply chips to a manufacturer that has not taken a patent licence was the specific conduct the Federal Trade Commission challenged and on which the Ninth Circuit reversed. The company continues it, and the legal position in the United States is that it is lawful, while regulators elsewhere have required modification.

Standard-essential patents. The broader argument concerns whether the commitment to license on fair, reasonable and non-discriminatory terms, which patent holders give to standards bodies in exchange for their technology being adopted, is meaningful when its interpretation is decided by litigation between parties of very unequal resources. Qualcomm's disputes with Apple, with Samsung, with Meizu and with others each turned on it.

Customer concentration and China. A very large share of Qualcomm's chip revenue comes from Chinese handset manufacturers, which exposes it directly to trade restrictions in both directions and to the Chinese policy of substituting domestic semiconductors. Chinese regulators opened an investigation in 2025 into Qualcomm's conduct in connection with an acquisition, a development read as at least partly responsive to American restrictions on Chinese technology.

Trade restrictions. Qualcomm has been affected by American controls on Huawei, which had been a customer, and by the broader decoupling, and its licences with Chinese manufacturers have been a persistent subject of negotiation.

How CDMA works

Qualcomm's history is difficult to follow without understanding what its founding technology actually does, and the explanation is not long.

Radio spectrum is finite, and the problem in mobile telephony is how many simultaneous conversations a given allocation can carry.

Frequency division gives each conversation its own channel, as analogue systems did. The number of conversations equals the number of channels, and the channel is occupied whether or not anyone is speaking.

Time division, used by GSM, divides each channel into repeating slots and allocates one to each user. This multiplies capacity by the number of slots and requires accurate synchronisation.

Code division abandons the idea of allocating anything. Every user transmits across the entire band at the same time, each multiplying their data by a distinct pseudorandom sequence far faster than the data itself, which spreads the signal across a wide bandwidth at low power. The receiver multiplies the received signal by the same sequence, which reconstructs the wanted data while the other users' signals, uncorrelated with that sequence, average toward noise.

The capacity of such a system is not fixed. It is determined by how much interference the receivers can tolerate, which means that adding a user degrades everyone slightly rather than being refused, and that anything reducing interference increases capacity directly. Three such things matter.

Voice activity detection. In a conversation each party speaks less than half the time. A variable-rate vocoder transmits at full rate during speech and a very low rate during silence, which reduces average interference by more than half and correspondingly increases the number of users.

Sectorisation. Directional antennas at the base station divide a cell into sectors, and users in one sector interfere less with those in another.

Soft handover. A handset near a cell boundary communicates with both base stations simultaneously and the network combines the results, which both improves reliability and reduces the transmit power required.

The problem that made engineers doubt the whole scheme is the near-far effect. A handset close to the base station arrives at a power many orders of magnitude greater than one at the cell edge, and swamps it entirely. The system therefore only works if every handset's signal arrives at approximately the same power, which requires the base station to command power adjustments to every handset hundreds of times a second with fractions of a decibel of precision, over a link that is itself subject to fading.

Qualcomm built that closed loop, together with the rake receiver, which exploits a property peculiar to spread spectrum: because the spreading sequence is fast, reflections arriving a fraction of a microsecond later are distinguishable from the direct path, so a receiver can track several of them and combine them constructively. Multipath, which degrades every other system, becomes a source of diversity gain.

The result was a system with several times the capacity of the alternatives, and a body of patents covering each of the mechanisms that make it work.

The licensing argument

The dispute over Qualcomm's business model is worth setting out in the terms the parties actually use, because it recurs in every jurisdiction and has never been resolved on the merits.

Qualcomm's position. Its research produced the system-level inventions that make a cellular device function, and those inventions are embodied in standards that every manufacturer implements. The value they create is realised in the complete device, not in any component, because a modem chip without the standard is worthless and a phone without the standard is not a phone. A royalty on the device price therefore reflects the contribution. Charging on the chip price would compensate the company for manufacturing a component rather than for inventing the system, and would make the very large research expenditure, approximately US$8 billion a year, impossible to justify. The company has also observed that it licenses at the device level to the manufacturer rather than to the chipmaker because that is where the standard is practised.

The manufacturers' position. A royalty on the device price means that improving a camera, a display or the storage increases what a manufacturer pays Qualcomm without Qualcomm contributing anything. The commitment Qualcomm gave to standards bodies, to license on fair, reasonable and non-discriminatory terms, was the condition on which its technology was adopted into the standard, and the position it now enjoys derives from that adoption rather than from competition. Refusing to license competing chipmakers, while requiring a licence as a condition of chip supply, uses the patent position to protect the chip business.

The regulators' positions have differed. China, South Korea, Taiwan and the European Union each found aspects of the conduct unlawful and imposed remedies, principally reductions in the royalty base, obligations to license, and prohibitions on exclusivity payments. The American Federal Trade Commission reached the same conclusion and was reversed on appeal, the Ninth Circuit holding that a firm has no general antitrust duty to deal with competitors, that a high royalty is a matter for patent and contract law rather than antitrust, and that the conduct was aggressive rather than exclusionary.

The state of the argument is therefore that the same facts have been assessed as unlawful in several jurisdictions and lawful in the company's home one, that the remedies imposed elsewhere have modified the model at the margin without changing its structure, and that the underlying question, what a holder of standard-essential patents may charge, remains among the least settled areas of technology law.

Reception

Qualcomm is regarded very differently by different constituencies, more so than most companies.

Among communications engineers its standing is extremely high. The CDMA work, the modem engineering, the millimetre wave implementation and the sustained contribution to standards are recognised as first-rate, and the company's research output and its engineers' presence in standards bodies are substantial.

Among its customers the relationship has been consistently adversarial. Apple, Samsung, Huawei, Meizu, BlackBerry, Nokia and others have each litigated or arbitrated with it, and several have invested heavily in alternatives specifically to reduce dependence.

Among regulators it has been among the most investigated technology companies in the world, with proceedings concluding against it on three continents.

Among investors the company has been valued at a persistent discount to comparable semiconductor businesses, on the reasoning that the licensing revenue is legally and politically vulnerable and that the chip business depends on a small number of customers who would prefer not to be customers. The diversification programme is directed as much at that perception as at the underlying exposure.

In San Diego the company is the defining institution of the local economy, having created a technology cluster in a city that had been dominated by defence and tourism, and the Jacobs family's philanthropy has shaped its universities and cultural institutions.

Assessment

Qualcomm is the clearest case in technology of a company whose value rests on a scientific bet made when nobody agreed with it.

The decision to build a commercial mobile system on code division multiple access was contested by essentially the entire telecommunications establishment on grounds that were technically serious rather than obtuse: the power control problem was real and the margin for error was very small. Qualcomm solved it, the capacity advantage proved genuine, and the consequence was that the company held essential patents in every subsequent generation of a technology that eight billion people now use.

The commercial arrangement it built on that position is the most contested in the industry and has now been tested by regulators in China, South Korea, Japan, Taiwan, the European Union and the United States, and by its largest customer in litigation across several countries. It has survived, principally because the Ninth Circuit held in 2020 that a company is not obliged to license its competitors and that vigorous exploitation of a lawfully obtained position is not an antitrust violation. Whether the outcome is desirable is a question about patent and standards policy rather than about Qualcomm, and it is not settled.

The present question is different and more ordinary. Qualcomm's core market is mature, its largest customer is replacing it, its Chinese customers are subject to trade policy, and its licensing revenue depends on a patent portfolio whose most valuable members will expire. The diversification into automotive, computing, connected devices and, most recently, data centre infrastructure is the response, and it is being conducted from a position of considerable strength, with approximately US$8 billion a year of research spending and a record of entering categories late and doing well in them.

The Snapdragon X series in personal computers is the clearest test. It is a technically strong product entering a market with entrenched incumbents and a software ecosystem built around a different instruction set, which is precisely the situation in which good hardware has repeatedly failed. The outcome will say more about the company's next decade than anything in its patent portfolio.