The Modem and the RadioNarrow moat

Qualcomm (QCOM) — moat facet

The hard part was never the silicon; it was every carrier network in every country signing off before the phone could ship.

For most of two decades the answer to the question of who made the best cellular modem was Qualcomm, and nobody serious argued about it.

QCT revenue streams, June 2026 quarter ($M)$5,086M -20%Handsets$1,830M +9%IoT$1,588M +61%AutomotiveTotal QCT revenue fell 5%. Two lines moving hard in opposite directions.
The diversification is real, and it is not yet bigger than the hole.

The difficulty is not the digital logic. It is that a modem has to work — reliably, at low power, on every carrier network in every country, across a dozen radio bands and four generations of standard, while the handset next to it is doing everything else. The certification burden alone runs to years of testing with operators who will not ship a phone that drops calls. Then it has to be integrated with the radio front end, the transceiver and the antenna tuning, because in a phone the parts are not separable in any meaningful sense.

That integration is where the franchise actually lives. Qualcomm sells a modem-to-antenna system, and the combined modem-and-application-processor Snapdragon carries higher revenue and margin than the standalone modem — a fact the company states plainly when explaining why its largest customer is its least profitable one.1 QCT turned $38.4 billion into $11.7 billion of pre-tax profit at a 30% margin on that basis.2

What Apple established is that this is a matter of money and time rather than of secrets. Having bought Intel's modem business in 2019, Apple shipped its own part in 2025 and is widening its use — and Qualcomm's handset revenue fell 20% year over year in the June 2026 quarter.3 Six years and several billion dollars, by the one customer with the volume to justify it.

The correct reading is not that the technology lead was fake. It is that a lead which takes six years to close is a narrow moat rather than a wide one, and that only a handful of buyers are large enough to fund the closing. Apple was the first. Samsung has had Exynos for years and keeps buying Snapdragon anyway, which is the counter-evidence.

The measure is Snapdragon content per premium Android handset, not modem share. If Qualcomm keeps selling the whole system where it used to sell the whole system, the franchise is intact whatever Apple does.

Moat trajectory: Narrowing

Apple has shipped its own modem and handset revenue fell 20% year over year. The technical lead is intact; the commercial position it supported is not, and the industry now knows what closing it costs.

The number that tests this moat
Reported
QCT pre-tax margin
30% — $11,670M on $38,367M

Up from 29% in fiscal 2024 on higher average selling prices. This is where concessions to large customers show up first, before they reach revenue: Qualcomm notes that share concentration among a few companies and their purchasing power may result in lower prices for its products. Watch the margin rather than the revenue line.

Source: Qualcomm Form 10-K, FY2025 ↗
Aspects of the moat
DIG DEEPER 01
Why a Modem Is Hard
Six years of certification relationships is a barrier that money alone buys slowly — which is exactly why only the largest customers try.
DIG DEEPER 02
Selling the Whole Chain, Not the Chip
A design win for the platform is worth a multiple of a design win for the modem, and it dies the moment one block goes in-house.
DIG DEEPER 03
No Fabs, and That Was the Right Answer
Spending on engineers instead of factories is why a $44 billion company earns forty-percent returns on capital.
DIG DEEPER 04
How Big the Lead Actually Was
Apple spent six years and several billion dollars to close it — which makes it a real moat and a finite one, and the distinction is the rating.
⚠ Threats to the moat
⚠ THREAT 01
The Customer Who Built the Replacement
Six years of co-engineering with the organisation that was designing you out of the product.
⚠ THREAT 02
MediaTek Does Not Have to Win, Only to Be Enough
MediaTek is not trying to beat Snapdragon. It is trying to move where premium starts, and 46% of Qualcomm's revenue is the buyer most willing to let…
References
  1. ReportedQualcomm sells a modem-to-antenna system, and the combined modem-and-application-processor Snapdragon carries higher revenue and margin than the standalone modem — a fact the company states plainly when explaining why its largest customer is its least profitable one.
    Qualcomm Incorporated, Form 10-K FY2025 — Item 1A, Risk Factors. Market-share concentration among a few companies, and the corresponding purchasing power of these companies, may result in lower prices for Qualcomm's products, which could adversely affect revenues and margins. Apple purchases Qualcomm's MDM (thin modem) products, which do not include its integrated application processor technology and which have lower revenue and margin contributions than the combined modem and application processor products; to the extent Apple devices using MDM products take share from customers using the combined products, revenues and margins would be adversely affected. The filing also carries risk factors on customers vertically integrating by developing their own integrated circuit products, on dependence on key personnel, and on the consequences of indebtedness. — FY2025 · publ. 2025-11-05 · source ↗
  2. ReportedQualcomm sells a modem-to-antenna system, and the combined modem-and-application-processor Snapdragon carries higher revenue and margin than the standalone modem — a fact the company states plainly when explaining why its largest customer is its least profitable one. QCT turned $38.4 billion into $11.7 billion of pre-tax profit at a 30% margin on that basis.
    Qualcomm Incorporated, Form 10-K FY2025 — Item 7, Management's Discussion and Analysis, and the reportable-segment note. QCT revenues $38,367M (2024 $33,196M, 2023 $30,382M) with EBT of $11,670M ($9,527M, $7,924M) at 30% of revenues (29%, 26%); QCT revenue streams handsets $27,793M ($24,863M, $22,570M), automotive $3,957M ($2,910M, $1,872M) and IoT $6,617M ($5,423M, $5,940M). QTL revenues $5,582M ($5,572M, $5,306M) against total costs and expenses of $1,539M ($1,545M, $1,678M), for EBT of $4,043M ($4,027M, $3,628M) — 72% of revenues in each of the three years. Reportable-segment revenues $43,949M, $38,786M and $35,716M against consolidated revenues of $44,284M, $38,962M and $35,820M. During the second quarter of fiscal 2025 Qualcomm executed final agreements for new long-term licensing arrangements. — FY2025 · publ. 2025-11-05 · source ↗
  3. ReportedHaving bought Intel's modem business in 2019, Apple shipped its own part in 2025 and is widening its use — and Qualcomm's handset revenue fell 20% year over year in the June 2026 quarter.
    Qualcomm Incorporated, Form 10-Q for the quarter ended 28 June 2026 (SEC, CIK 804328). Revenues $9,947M for the quarter against $10,365M a year earlier, and $32,798M for the nine months against $33,013M; equipment and services $8,475M ($8,893M) and licensing $1,472M ($1,472M). Cost of revenues $4,670M, research and development $2,607M ($2,226M), selling, general and administrative $976M ($771M). Income before income taxes $2,462M ($2,952M); income tax expense $460M for the quarter, and a $4,136M income tax benefit for the nine months against a $1,034M expense — net income $2,002M ($2,666M) for the quarter and $12,377M ($8,658M) for the nine months. Diluted earnings per share $1.87 ($2.43) on 1,069 million diluted shares (1,099). QCT revenue streams for the quarter: handsets $5,086M ($6,328M), automotive $1,588M ($984M), IoT $1,830M ($1,681M), total QCT $8,504M ($8,993M). — Q3 FY2026 · publ. 2026-07-29 · source ↗
Sources
Generated September 23, 2026