The Tax on Every Pocket
In any given second, somewhere on Earth, a Qualcomm chip is negotiating a handshake with a cell tower — modulating a radio signal into the precise mathematical pattern that lets a stream of ones and zeros traverse the atmosphere at the speed of light and arrive, intact, as a TikTok video, a wire transfer, a 911 call. The company does not make the phones that most people hold, does not operate the networks those phones connect to, does not write the software that lights up their screens. And yet Qualcomm extracts a royalty — typically 3.25% to 5% of the wholesale price — on virtually every multimode smartphone sold anywhere in the world, whether or not that phone contains a single Qualcomm transistor. This is the most elegant and contentious business model in the history of semiconductors: a licensing structure so deeply embedded in the global wireless standard that attacking it has consumed the legal budgets of Apple, Huawei, the Federal Trade Commission, the Korean Fair Trade Commission, the European Commission, and the Chinese National Development and Reform Commission, and yet the structure stands. Qualcomm's entire existence is an argument — still being litigated, still being won — that the most durable monopoly is not a product monopoly but a standards monopoly, and that the most valuable real estate in technology is not a fab or a data center but a portfolio of patents declared essential to the way the modern world communicates.
The numbers tell a story of almost surreal capital efficiency. In fiscal year 2024, Qualcomm generated approximately $38.96 billion in revenue. Its QTL (Qualcomm Technology Licensing) division — the unit that collects royalties on those essential patents — produced roughly $5.6 billion in revenue on margins that approach 70%, an annuity stream unmatched in the chip industry. The QCT (Qualcomm CDMA Technologies) division, which designs and sells the Snapdragon system-on-chips that power the majority of the world's premium Android phones, contributed the bulk of the remaining revenue. Together, these two businesses create a flywheel of unusual structural power: profits from licensing fund the R&D that produces the next generation of chips and patents, which in turn refresh the licensing portfolio. The flywheel has been spinning since 1985, through at least four generational transitions in wireless technology — from analog to CDMA to 3G to 4G to 5G — and each turn has made Qualcomm harder to dislodge.
By the Numbers
Qualcomm at a Glance
$38.96BFY2024 revenue
~$5.6BQTL licensing revenue (FY2024)
~70%QTL operating margin
$188B+Approximate market capitalization (early 2025)
~51,000Employees worldwide
140,000+Active patent portfolio (est.)
5Generations of wireless standard leadership (CDMA → 5G)
3.25–5%Typical per-device licensing royalty rate
The paradox embedded in these numbers is the paradox that has defined Qualcomm for four decades. The company that invented CDMA — the multiplexing technique that made modern mobile communication possible — has always been more valuable as a legal entity than as a silicon entity, even as its silicon has become among the best in the world. Qualcomm's chip business is enormous and growing. But the licensing business is the one that warps the competitive landscape around it. Rivals design around Qualcomm's chips routinely. No one has successfully designed around Qualcomm's patents.
The Professor's Gambit
Irwin Mark Jacobs was not the archetype of a Silicon Valley founder. He was a professor — MIT-trained, a specialist in information theory and digital communications who co-authored a definitive textbook on the subject before he co-authored a company. Born in 1933 in New Bedford, Massachusetts, the son of a restaurant owner, Jacobs had the gift of translating
Claude Shannon's elegant mathematics into systems that worked in the physical world, where signals bounce off buildings and interfere with each other in ways that textbooks handle with abstractions and engineers handle with ulcers. By the time he co-founded Qualcomm in July 1985, in a room above a San Diego pizza restaurant with six co-founders and a name that stood for "
Quality Communications," Jacobs was already 52 years old and had already built and sold one company — Linkabit, a defense communications firm he'd founded in 1968.
The founding team — Jacobs, Andrew Viterbi (inventor of the Viterbi algorithm, a cornerstone of digital communication), Harvey White, Adelia Coffman, Andrew Cohen, Klein Gilhousen, and Franklin Antonio — possessed an improbable concentration of theoretical firepower. Viterbi alone would have anchored most research labs. Gilhousen, a quiet engineer obsessed with spread-spectrum radio, was arguably the person most responsible for making CDMA work in practice. These were not entrepreneurs in the contemporary sense. They were scientists who happened to believe that the telecommunications industry had made a catastrophic technical error, and that correcting it could be worth billions.
The error was this: the global wireless industry, in the late 1980s, was converging on a standard called TDMA (Time
Division Multiple Access) for second-generation cellular networks. TDMA divided each radio channel into time slots, giving each user a sliver of time on a shared frequency. It was incremental, comprehensible, and — Qualcomm's founders believed — fundamentally limited. CDMA (Code Division Multiple Access) was the radical alternative: instead of dividing time or frequency, it let every user transmit simultaneously on the same frequency, distinguishing them through unique mathematical codes. It was a technique borrowed from military spread-spectrum radio, and the established wisdom in telecommunications held that it could never work at commercial scale. The interference problem — every user's signal appearing as noise to every other user — seemed intractable.
Jacobs and his team set out to prove the industry wrong. Early demonstrations in the late 1980s and early 1990s were theatrical and high-stakes: Qualcomm would invite skeptical executives from carriers and equipment manufacturers to witness live CDMA calls, and the technology would perform — sometimes. The demonstrations were, by several accounts, triumphs of careful engineering and selective showmanship. But the results were real. CDMA could carry roughly ten times more calls per unit of spectrum than the AMPS analog systems it aimed to replace, and significantly more than TDMA. Spectrum is the scarcest resource in wireless — every regulator on Earth auctions it for billions — and a technology that multiplied its effective capacity was, if it worked, the equivalent of discovering new radio spectrum.
The political battle was as fierce as the technical one. European carriers and Ericsson had bet heavily on GSM, a TDMA-based standard. The American carrier landscape was split. Qualcomm needed at least one major carrier to adopt CDMA to prove commercial viability. That carrier turned out to be what was then called PacTel — later subsumed into what became Verizon — and the Korean carriers, who saw CDMA as a national strategic bet that would free Korea from dependence on European and Japanese technology. South Korea's adoption of CDMA in the mid-1990s, championed by the government and executed by Samsung and LG, was one of the pivotal moments in both Qualcomm's history and the trajectory of the Korean electronics industry. It gave CDMA a massive, dense, technically demanding market in which to prove itself — and it worked.
We were told by virtually everyone in the industry that CDMA would never work. The physics was against us, they said. We had the physics on our side.
The Licensing Cathedral
What happened next is the move that made Qualcomm, Qualcomm. Most semiconductor companies that develop a breakthrough technology monetize it by selling chips. Qualcomm did that too — eventually. But first, Jacobs and his team did something far more consequential: they embedded their patents into the standard itself.
The mechanics are technical but the strategic logic is breathtakingly simple. Wireless communication requires interoperability — your phone must be able to talk to any tower, anywhere. This demands standards, set by bodies like the ITU, 3GPP, and their predecessors. When a technology is adopted into a standard, every device that implements that standard must use the patented technique. The patent holder is required to license on FRAND terms — Fair, Reasonable, and Non-Discriminatory — but the precise meaning of "reasonable" has been the subject of more litigation than perhaps any other word in the English language.
Qualcomm's CDMA patents were not merely useful to the standard. They were the standard. The company held patents on the fundamental waveform, the power control mechanisms, the soft handoff techniques, the encoding and decoding processes that made CDMA function. When CDMA was adopted as the basis for 3G (as WCDMA and CDMA2000), Qualcomm's patent portfolio became essential to every 3G device on the planet — including, crucially, GSM-evolved devices that used WCDMA, meaning that even the European carriers who had fought CDMA were now paying royalties to Qualcomm.
The licensing model that Jacobs established charged royalties not on the price of the modem or the baseband chip, but on the wholesale price of the entire device. A smartphone selling for $500 at wholesale might generate $16 to $25 in Qualcomm licensing fees — regardless of whether that phone used a Qualcomm chip or a competitor's. This was the detail that drove competitors to apoplexy. As phones became more expensive — adding cameras, screens, processors, memory — Qualcomm's per-unit royalty grew, even though the communications technology itself was a diminishing fraction of the device's value. The model made Qualcomm the de facto tax collector of the mobile industry.
How standard-essential patents create inescapable licensing obligations
Every company that contributes technology to a wireless standard must commit to licensing its standard-essential patents (SEPs) on FRAND terms. But FRAND contains a structural ambiguity: "reasonable" royalty rates are never pre-defined. The patent holder proposes; the implementer negotiates; courts resolve disagreements. Qualcomm's position — holding the largest portfolio of SEPs for 3G, 4G, and 5G — gives it enormous leverage in every negotiation. The alternative to paying Qualcomm's rate is not "use a different technology." The alternative is "don't build a phone."
The model was, and is, uniquely profitable. QTL requires almost no marginal cost to operate — no fabs, no supply chain, no inventory risk. Every new phone sold anywhere in the world, by any manufacturer, triggers a royalty payment to San Diego. The business scales with the global smartphone market without requiring Qualcomm to design, manufacture, or ship a single additional product. In fiscal years when QTL has operated at peak efficiency, its operating margins have exceeded 70% — a figure that would be extraordinary for a software company, let alone one embedded in the hardware supply chain.
But the model also made Qualcomm the most sued technology company of its era. The list of legal adversaries reads like a directory of global regulatory power.
The War of All Against Qualcomm
The attacks came in waves, and they came from everywhere.
Korea's Fair Trade Commission hit Qualcomm with a $208 million fine in 2009 for alleged anticompetitive licensing practices. China's National Development and Reform Commission followed in 2015 with a $975 million fine — at the time, the largest antitrust penalty ever levied in China — settling an investigation into whether Qualcomm's royalty rates for Chinese manufacturers were unreasonably high. Qualcomm agreed to lower its royalty rates in China, a concession that reverberated through its financial statements for years. The European Commission opened its own investigations. Taiwan's Fair Trade Commission fined the company $773 million in 2017.
Then came the existential battle: Apple.
In January 2017, Apple sued Qualcomm for approximately $1 billion in allegedly overcharged royalties, arguing that Qualcomm's licensing model was coercive, anticompetitive, and built on a refusal to license its patents to rival chipmakers directly. Apple alleged that Qualcomm withheld rebates to punish Apple for cooperating with Korean regulators. Qualcomm countersued. The litigation cascaded across jurisdictions — U.S. federal courts, Chinese courts, German courts, the International Trade Commission. Apple stopped paying royalties. Qualcomm's revenue cratered by billions. The FTC, under Chairwoman Edith Ramirez and then Makan Delrahim (whose tenure introduced complications), filed its own antitrust suit against Qualcomm in January 2017, echoing many of Apple's arguments.
The FTC case, FTC v. Qualcomm, produced a dramatic first-act victory for the government. In May 2019, Judge Lucy Koh of the Northern District of California issued a sweeping ruling finding that Qualcomm's licensing practices violated antitrust law — ordering the company to renegotiate its licenses, license its patents to rival chipmakers, and submit to seven years of compliance monitoring. The ruling sent Qualcomm's stock plunging and prompted speculation that the company's licensing model was finished.
It wasn't. In August 2020, the Ninth Circuit Court of Appeals reversed Judge Koh's decision in a 3-0 opinion that reads as a near-total vindication of Qualcomm's business model. The appellate court found that Qualcomm had no obligation to license its patents to rival chipmakers, that its royalty rates did not constitute an anticompetitive surcharge, and that the district court had fundamentally misapplied antitrust law. The FTC did not appeal to the Supreme Court.
The Apple dispute resolved two months before the Ninth Circuit ruling, in April 2019, with a settlement whose terms were confidential but reportedly included a payment from Apple to Qualcomm of approximately $4.5 billion and a six-year licensing agreement through at least 2025, with a two-year option to extend. Apple simultaneously signed a multiyear chipset supply agreement, returning Qualcomm modems to iPhones after a brief period using Intel modems. Intel, for its part, exited the smartphone modem business entirely within days of the settlement — a capitulation that underscored just how difficult it is to compete in baseband without Qualcomm's patent portfolio as both sword and shield.
Our technology is foundational to the connected world. Licensing reflects the value of that foundation.
The settlement with Apple was a defining moment. Qualcomm had survived simultaneous legal assault from the world's most valuable company, the world's most powerful antitrust regulator, and a half-dozen sovereign competition authorities — and emerged with its licensing model intact, its royalty rates only modestly reduced, and its largest customer back in the fold. The market understood what this meant. Qualcomm's stock more than doubled from its 2019 lows within two years.
The Snapdragon Doctrine
If licensing is Qualcomm's cathedral, the chip business is its fortress — and for the past decade, that fortress has been expanding aggressively into territory far beyond the smartphone.
The QCT division designs system-on-chips (SoCs) — integrated processors that combine a CPU, GPU, AI accelerator, image signal processor, and cellular modem into a single piece of silicon. The flagship Snapdragon 8 series has been the premium Android SoC of choice for most of the past decade, powering devices from Samsung, Xiaomi, OnePlus, and dozens of other manufacturers. In fiscal year 2024, QCT generated approximately $33 billion in revenue — the vast majority from handsets, but with rapidly growing contributions from automotive, IoT (Internet of Things), and PC processors.
The strategic logic of QCT is inseparable from QTL. Qualcomm's modem technology — the component that connects a device to cellular networks — is the one piece of silicon that no competitor has been able to replicate at Qualcomm's performance level. The company's modem IP is the direct descendant of its CDMA work, refined through five generational transitions, and it integrates seamlessly with the Snapdragon application processor. This integration creates a bundling advantage: a phone manufacturer choosing a competitor's application processor (say, MediaTek's Dimensity) must either use that competitor's inferior modem or attempt the engineering nightmare of pairing a Qualcomm modem with a non-Qualcomm processor. Most choose the integrated Qualcomm solution.
MediaTek, the Taiwanese chipmaker, has emerged as Qualcomm's most formidable competitor in mobile SoCs, particularly in the mid-range and low-end segments that dominate global smartphone volume. MediaTek's Dimensity chips have gained significant share in China, India, and Southeast Asia, pressuring Qualcomm's QCT margins. But in the premium tier — devices selling for $400 and above — Qualcomm's Snapdragon 8 Gen 3 and its successors retain a dominant position, owing largely to the modem integration advantage and to sustained investment in the CPU and GPU architectures (Qualcomm's custom Kryo and Adreno designs, with the more recent Oryon CPU core representing a significant in-house architectural effort).
🔧
The Snapdragon Ecosystem
How Qualcomm's chip portfolio has expanded beyond mobile
2013Snapdragon becomes the dominant premium Android SoC.
2016Qualcomm enters automotive with Snapdragon Ride platform.
2018Launch of Snapdragon Compute for always-connected PCs.
2021Acquires Nuvia for $1.4 billion to develop custom CPU cores.
2023Snapdragon X Elite laptop chip announced, targeting Apple's M-series.
2024Automotive design-win pipeline reportedly exceeds $45 billion.
The Nuvia acquisition deserves attention. Nuvia was founded in 2019 by Gerard Williams III, who had been Apple's chief CPU architect — the engineer behind the A-series chips that powered the iPhone and, eventually, the M-series chips that powered the Mac. When Qualcomm acquired Nuvia in March 2021 for $1.4 billion, it was buying not just a team but an architectural philosophy: the belief that custom-designed CPU cores, optimized from the ground up rather than licensed from Arm's standard designs, could deliver step-function improvements in performance per watt. The Snapdragon X Elite, announced in late 2023 and shipping in mid-2024, is the first product of this effort — a laptop processor that Qualcomm positioned as a direct competitor to Apple's M3, with AI inference capabilities that Intel and AMD had not yet matched in the PC form factor. (Apple sued Qualcomm over the Nuvia acquisition, alleging that Arm-based technology licenses held by Nuvia employees did not transfer to Qualcomm. That dispute, too, was resolved in Qualcomm's favor at trial in December 2024.)
The Man After Jacobs
The transition from Irwin Jacobs to Paul Jacobs (his son) to Steve Mollenkopf to Cristiano Amon traces an arc from inventor-founder to professional manager to operational executor to expansionist CEO — each succession reflecting the company's strategic priorities in its era.
Irwin Jacobs stepped down as CEO in 2005 and as chairman in 2009, having built the cathedral. Paul Jacobs, who became CEO in 2005 at age 43, navigated the 3G-to-4G transition and expanded QCT into the smartphone era — but also oversaw a period of increasing antitrust scrutiny and the early stages of the conflict with Apple. Steve Mollenkopf, a career Qualcomm engineer who became CEO in 2014, inherited the legal maelstrom and steered the company through the Apple lawsuit, the FTC case, and the failed hostile takeover attempt by Broadcom in 2017–2018.
That Broadcom episode is worth lingering on. In November 2017, Broadcom CEO Hock Tan — a legendary dealmaker with a reputation for acquiring companies and cutting costs to the bone — made an unsolicited offer to acquire Qualcomm for $105 billion, later raised to $121 billion, which would have been the largest technology acquisition in history. Qualcomm's board unanimously rejected the bid, arguing that it undervalued the company and that Broadcom's cost-cutting philosophy would destroy Qualcomm's R&D engine. The battle escalated to the point where Broadcom attempted to replace Qualcomm's entire board at the 2018 annual meeting. President Trump intervened in March 2018, issuing an executive order blocking the deal on national security grounds — a move recommended by the Committee on Foreign Investment in the United States (CFIUS), which cited Broadcom's Singapore incorporation and the risk that a Broadcom-owned Qualcomm would underinvest in 5G research, ceding American leadership to Huawei.
The presidential intervention was extraordinary and revealing. It signaled that the U.S. government viewed Qualcomm not merely as a chipmaker but as a national strategic asset — the company whose 5G patents and technology represented America's best hope of maintaining leadership in the next generation of wireless infrastructure. The geopolitical dimension of Qualcomm's business — always present in the background, given the company's defense origins and its centrality to telecommunications — had become explicit.
Cristiano Amon, a Brazilian-born engineer who had risen through QCT and was widely regarded as the architect of Qualcomm's 5G chip strategy, became CEO in June 2021. Amon's Qualcomm is defined by diversification: the deliberate, aggressive expansion of the chip business beyond smartphones and into automotive, PCs, industrial IoT, and edge AI. His stated goal is to reduce Qualcomm's dependence on the handset market — which is mature, cyclical, and increasingly contested by MediaTek — by building large-scale businesses in adjacent markets where Qualcomm's modem-plus-processor integration and AI capabilities provide a structural advantage.
We are no longer a wireless communications company. We are a connected computing company for the intelligent edge.
The 5G Harvest
Each generational transition in wireless technology has been an existential bet for Qualcomm and, simultaneously, its greatest source of value creation. The pattern is remarkably consistent: Qualcomm invests billions in R&D during the pre-standard phase, contributes foundational technology to the standard-setting process, accumulates standard-essential patents, and then harvests royalties for the fifteen-to-twenty-year life of that standard generation.
5G is the latest and largest iteration of this pattern. Qualcomm's 5G patent portfolio is estimated to be the largest in the world, encompassing fundamental contributions to 5G NR (New Radio) in both sub-6 GHz and millimeter-wave frequencies. The company was the first to demonstrate working 5G modem chipsets, the first to ship commercial 5G modems at scale (the Snapdragon X50, then X55, then X60, X65, X70, and X75), and the first to integrate 5G modems into system-on-chips for smartphones. This first-mover position in 5G silicon translated directly into market share gains — Qualcomm supplied the 5G modem for the initial wave of 5G phones from Samsung, Xiaomi, Oppo, and even (after the Apple settlement) Apple's iPhone 12 and subsequent models.
The 5G transition also expanded Qualcomm's addressable market beyond phones. 5G's promise of ultra-low latency and massive device density made it relevant to automotive (vehicle-to-everything communication), industrial automation, fixed wireless access (replacing home broadband with cellular), and private network deployments. Qualcomm's 5G modem technology became the connective tissue for these new markets, giving the QCT diversification strategy a technical foundation rather than a merely aspirational one.
But the 5G harvest has been slower than some investors hoped. The global 5G rollout has proceeded unevenly — fast in China, South Korea, and parts of the U.S., slower in Europe and much of the developing world. The "killer app" for 5G — the use case that makes consumers eager to upgrade — has not materialized with the urgency that the industry projected. Smartphone replacement cycles have lengthened. And the premium Qualcomm commands for 5G-capable chips has compressed as MediaTek and Samsung's Exynos division have brought their own 5G solutions to market.
Cars, Laptops, and the Edge
The automotive push is perhaps the most consequential strategic bet of the Amon era. Qualcomm's Snapdragon Digital Chassis platform — an integrated suite of chips for infotainment, advanced driver-assistance systems (ADAS), vehicle-to-cloud connectivity, and digital cockpit displays — has accumulated a design-win pipeline that the company values at over $45 billion. Customers include General Motors, BMW, Mercedes-Benz, Hyundai, Renault, and Stellantis. The Snapdragon Ride platform targets autonomy. The revenue cycle in automotive is glacially slow compared to smartphones — design wins signed today may not generate meaningful revenue for three to five years — but the per-vehicle silicon content is far higher than in a phone, and automotive customers, once committed to a platform, rarely switch mid-cycle.
The PC initiative centers on the Snapdragon X Elite and Snapdragon X Plus processors, launched in partnership with Microsoft for Windows on Arm. The pitch is compelling: Apple's M-series chips proved that Arm-based processors could deliver superior performance per watt compared to x86 architecture, and Qualcomm — with the Nuvia team's custom Oryon cores — aims to bring a similar advantage to the Windows ecosystem. Early reviews of Snapdragon X Elite laptops in 2024 were mixed but promising: exceptional battery life and AI inference performance, but application compatibility issues stemming from the x86-to-Arm translation layer. Whether Qualcomm can crack the PC market at scale depends on factors partly outside its control — Microsoft's commitment to Arm optimization, the ISV ecosystem's willingness to port applications natively, and the inertia of the x86 installed base.
The IoT business, while less glamorous, is substantial — several billion dollars in annual revenue across industrial, consumer, and networking applications. Qualcomm's Wi-Fi chipsets are embedded in routers, mesh systems, and enterprise access points from virtually every major networking vendor. The company's Bluetooth and ultra-wideband chips power earbuds, smartwatches, and asset trackers. This business lacks the narrative clarity of the automotive or PC efforts, but it generates consistent revenue and extends Qualcomm's reach into the broader connected device ecosystem.
🚗
Qualcomm's Diversification Math
Revenue mix shift targets
| Segment | FY2024 Revenue (est.) | Target by ~FY2029 |
|---|
| Handsets (QCT) | ~$24B | Stable / modest growth |
| Automotive (QCT) | ~$2.9B | $9B+ (pipeline-driven) |
| IoT (QCT) | ~$5.4B | $7–9B |
| PC / Compute (QCT) | Nascent | Meaningful contributor |
| QTL (Licensing) | ~$5.6B | ~$5–6B (stable) |
The Apple Question
No single relationship defines Qualcomm's near-term trajectory more than Apple. And no single relationship is more ambiguous.
Apple has been developing its own cellular modem in-house since at least 2018, when it poached engineers from Intel and Qualcomm to build a team in San Diego (literally across the street from Qualcomm's campus, in a move that carried the unmistakable aroma of corporate spite). The effort has been plagued by delays. Apple's modem was initially expected to appear in iPhones by 2023, then 2024, then 2025. Reports in late 2024 suggested that Apple's first in-house modem would debut in a limited capacity — possibly in a lower-end iPhone SE model — with full integration into flagship iPhones still years away. The technical challenge is immense: building a modem that supports every cellular standard across every frequency band in every country, while meeting the power efficiency and heat dissipation requirements of a smartphone, is arguably the most difficult engineering problem in consumer electronics. Qualcomm has been doing it for thirty years. Apple is starting from scratch.
The strategic implications are stark. If Apple successfully internalizes its modem, Qualcomm loses its single largest chip customer — Apple is estimated to account for roughly 20–25% of QCT revenue. The licensing impact is more contained; Apple would still owe royalties on Qualcomm's standard-essential patents regardless of whose modem is inside the iPhone. But the chip revenue loss would be significant, and the signal it sends to other customers (Samsung, which already uses its own Exynos modems in some markets) could accelerate defections.
Qualcomm has been preparing for this scenario. The diversification into automotive, PC, and IoT is partly an insurance policy against Apple's departure. Amon has publicly stated that Qualcomm's financial model does not assume retaining Apple as a chip customer beyond the current agreement period. The company has been managing investor expectations while simultaneously executing to make the non-Apple businesses large enough to absorb the blow.
The uncertainty is the point. Qualcomm is simultaneously the indispensable supplier to Apple and the company Apple most wants to displace. The relationship is a loaded spring — and the timing of its release will shape Qualcomm's revenue trajectory for the rest of the decade.
The Geopolitical Modem
Qualcomm occupies a unique position at the intersection of American technology dominance and Chinese manufacturing might. The company derives a substantial portion of its revenue from Chinese OEMs — Xiaomi, Oppo, Vivo, and others — and from devices sold in the Chinese market. At the same time, Qualcomm's technology is subject to U.S. export controls that have constrained its ability to sell advanced chips to Huawei, the Chinese national champion in telecommunications.
The Huawei situation is a case study in how geopolitics can simultaneously create and destroy value. Before the U.S. Commerce Department placed Huawei on the Entity List in May 2019, Huawei was one of Qualcomm's largest customers. The export ban severed that relationship for advanced chips, though Qualcomm obtained limited licenses to sell 4G (but not 5G) chipsets to Huawei for certain products. Huawei's response was to develop its own HiSilicon Kirin chips with in-house modem technology — a process accelerated by desperation and subsidized by the Chinese government. In 2023, Huawei shocked the industry by shipping the Mate 60 Pro with an apparently advanced 5G chip manufactured by SMIC on a 7nm-class process, suggesting that Chinese semiconductor capabilities were further advanced than Western intelligence had estimated.
For Qualcomm, the geopolitical chessboard presents a dual risk: the ongoing possibility of further U.S. export restrictions that could limit sales to Chinese customers, and the accelerating development of indigenous Chinese chip capabilities that could reduce Chinese OEMs' dependence on Qualcomm over time. The company has navigated this terrain with unusual diplomatic skill — maintaining relationships in Beijing while cooperating with Washington's security concerns — but the structural trajectory is toward greater fragmentation of the global semiconductor supply chain, and Qualcomm sits squarely on the fault line.
AI at the Edge
Qualcomm's AI narrative is distinct from the data center AI story that has driven Nvidia's parabolic valuation. Qualcomm is not competing for the training workloads that require thousands of GPUs in a hyperscaler's data center. Instead, Qualcomm is betting on inference at the edge — the execution of AI models directly on devices, without sending data to the cloud.
The logic is sound, both technically and commercially. Running AI locally on a phone, a car, a laptop, or an industrial sensor avoids the latency and privacy concerns of cloud inference, reduces bandwidth costs, and enables functionality in disconnected environments. Qualcomm's Snapdragon processors have included dedicated AI accelerators (the Hexagon NPU) since 2017, and the company claims that its latest Snapdragon 8 Gen 3 delivers over 45 TOPS (tera-operations per second) of AI inference performance. The Snapdragon X Elite laptop chip was explicitly positioned as an "AI PC" processor, with NPU performance designed to run large language models and generative AI applications locally.
Whether on-device AI becomes the transformative use case Qualcomm needs — or remains a feature bullet point on spec sheets — is one of the most consequential open questions in the company's near-term outlook. Qualcomm has the hardware advantage: its NPU technology is several years ahead of what Intel and AMD ship in comparable form factors. But the software ecosystem — the availability of compelling AI applications optimized for on-device inference — is still developing. If the AI future is mostly cloud, Qualcomm's edge AI bet is a feature, not a platform. If the AI future is hybrid, with meaningful computation happening on devices, Qualcomm's position is formidable.
The Cathedral and the Fortress
Step back and the architecture of Qualcomm's business reveals itself as two interlocking structures, each reinforcing the other but each with its own logic, its own vulnerabilities, its own timescale.
The cathedral is QTL — the licensing business built on thirty-nine years of R&D, on patents declared essential to every generation of wireless standard from CDMA to 5G, on legal precedents won in courtrooms across four continents. It is the most profitable recurring revenue stream in semiconductors. It requires almost no incremental capital. And it faces a slow, structural challenge: as wireless standards evolve, new contributors — Samsung, Huawei, Nokia, Ericsson — accumulate their own essential patents, diluting Qualcomm's share of the total SEP pool. Qualcomm's share of 5G essential patents, while still the largest of any single company, is smaller proportionally than its share of 3G patents. The licensing moat is not widening. It is being maintained at enormous legal and R&D expense.
The fortress is QCT — the chip business that designs the silicon powering billions of connected devices. It is capital-intensive (Qualcomm spends approximately $8–9 billion annually on R&D), subject to the brutal cyclicality of consumer electronics demand, and facing intensifying competition from MediaTek below, Apple within, and potentially from new entrants enabled by Arm's increasingly flexible licensing model. But it is also the vehicle for Qualcomm's diversification — the business that can grow into automotive, PCs, IoT, and industrial markets where the licensing model alone cannot reach.
The genius of the structure is that the cathedral funds the fortress and the fortress defends the cathedral. Licensing profits finance the R&D that produces the next generation of patentable innovations. Chip leadership ensures that Qualcomm remains at the frontier of wireless technology, which in turn ensures that its engineers are the ones contributing the most essential inventions to each new standard. The danger is that the cycle could reverse: if chip share erodes and R&D spending must be cut, patent output declines, licensing leverage weakens, profits shrink, and the flywheel decelerates.
The company sitting in San Diego, across the street from the Pacific Ocean, in a campus that has expanded to accommodate tens of thousands of engineers, still runs on the insight that Irwin Jacobs had in 1985: that the most valuable position in wireless is not making the hardware or operating the network but owning the mathematics. The mathematics, encoded in patents and silicon, is what converts the invisible spectrum into human connection. And Qualcomm, for nearly four decades, has been the company that converts human connection into cash flow — reliably, litigiously, brilliantly.
The latest Snapdragon chip sits in a device in someone's pocket right now, negotiating a handshake with a tower, modulating a signal, paying a tax. The device's owner will never know. That is the design.
