The Modem-to-Antenna War.Original analysisNot investment advice
Qualcomm wants to sell modem-to-antenna. Apple wants to own device-to-network. The iPhone 12’s USI mmWave antenna module was an early crack. Apple’s C1 modem makes the strategy clearer.
In 2020, the iPhone 12 teardown revealed a small but important fracture in Qualcomm’s 5G moat. Apple was using Qualcomm’s 5G modem system, but not Qualcomm’s complete mmWave antenna module. The iPhone 12 mmWave antenna modules came from USI, a Shanghai-based system-in-package design and manufacturing supplier.12
At first glance that looked like normal Apple supplier behaviour. Diversify suppliers. Squeeze cost. Avoid single-vendor dependence. The deeper signal mattered more. Qualcomm did not just want to sell a modem. Qualcomm wanted to sell the full modem-to-antenna chain, where the baseband, transceiver, RF front-end, antenna tuners, mmWave modules, and software all worked better as one integrated system. Apple did not want to buy the system blindly. The iPhone 12 showed that Apple could keep the Qualcomm modem while replacing pieces around it.
In 2026, that story looks much bigger. Apple now ships the iPhone 16e with the Apple C1 cellular modem. The official spec sheet lists 5G sub-6 GHz with 4×4 MIMO. It does not list mmWave.3 That detail is the whole story in one bullet point.
Apple has crossed the first bridge. It can ship its own cellular modem in a real iPhone. It has not yet crossed the hardest bridge: full high-end global modem-RF parity with Qualcomm, especially mmWave.
This is the modem-to-antenna war.
The correct claim is not that Qualcomm lost the modem war. The correct claim is that Apple is unbundling Qualcomm’s modem-RF moat. Qualcomm still has the best complete engine. Apple is trying to own the car.
I. The 2020 thesis
In October 2020, Dylan Patel published a SemiAnalysis piece on the iPhone 12 mmWave antenna decision. The piece argued that even though Apple used Qualcomm’s X55 5G modem-RF system, Apple chose USI antenna modules for mmWave, not Qualcomm’s own integrated modules.1 Apple’s thin iPhone industrial design likely pushed it toward a more custom mmWave antenna path. The implication was structural. Qualcomm’s modem leadership did not automatically protect every layer of the RF stack around it.
I revisited that piece because the small antenna decision turned out to be a leading indicator. Apple kept buying Qualcomm modems for years after the iPhone 12. But the unbundling continued.
Qualcomm’s modem leadership does not automatically protect every layer of the RF stack if Apple can replace components around the modem.
Six years later, Apple has done exactly that, just not all at once and not at full parity. The iPhone 16e is the first iPhone that shipped without a Qualcomm modem at all, and the spec line that matters is what Apple did and did not commit to.
II. A smartphone radio is not one chip
The reason this matters is that a modern cellular phone is not a modem. It is a chain. The baseband chip processes the cellular protocol. The RF transceiver converts between baseband and radio frequencies. The RF front-end pushes signal through amplifiers, filters, and switches. Antenna tuners adapt the antenna to changing conditions. mmWave antenna modules and the antennas themselves close the link to the carrier network. Underneath that chain sits power, thermal design, board layout, software, and global carrier certification.
Qualcomm’s strength is integration. Apple’s strength is control.
III. Why mmWave is hard
mmWave uses higher frequencies than sub-6 GHz. It can deliver large peak bandwidth, but it has shorter range, weaker penetration through buildings and bodies, and difficult antenna placement. Phones need multiple antenna modules placed around the device to handle blockage. That makes mmWave a packaging, antenna, RF, thermal, and industrial design problem, not only a modem problem.
The market is also uneven. Apple lists 5G sub-6 GHz and mmWave for its US iPhone 17 specs.7 Apple’s Australian iPhone 17 specs list 5G sub-6 GHz but not mmWave.8 The same handset family, the same brand, different RF surface area, because the network is different.
The GSA July 2025 mmWave report tells the same story in aggregate. 203 operators across 56 countries and territories are investing in 5G mmWave, but only 24 operators in 17 countries had launched 5G networks using mmWave spectrum.11 Sub-6 GHz is the mainstream global layer. mmWave is the region-specific peak.
The global mainstream
- Range · long, penetrates buildings.
- Bandwidth · high but not extreme.
- Antennas · conventional placement.
- Deployment · broad, almost every market.
The regional peak
- Range · short, blocked by bodies and walls.
- Bandwidth · very high, low latency at line-of-sight.
- Antennas · multiple modules around the device.
- Deployment · concentrated in select markets.
IV. Apple’s C1 proves direction, not full parity
The iPhone 16e is the first iPhone in history that does not contain a Qualcomm modem. Apple’s official tech specs list the Apple C1 cellular modem and 5G sub-6 GHz with 4×4 MIMO. They do not list mmWave.3
Reuters reported that the C1 is Apple’s first custom modem subsystem and is positioned to reduce Apple’s reliance on Qualcomm, with the chip itself fabricated using 4nm and 7nm process technologies and tested across approximately 180 carriers in 55 countries during development.4 Reading that alongside Apple’s own spec sheet, the message is consistent. Apple can ship its own modem. Apple is not yet shipping it in the highest-end iPhone or with the hardest RF tier.
Apple is starting where the risk is lower. Sub-6 first, mmWave later.
V. Qualcomm is still the technical leader
This section needs to be honest, because the temptation to write a Qualcomm-loses narrative is high, and the data does not support it.
In September 2023, Qualcomm and Apple extended their modem agreement so that Qualcomm would supply Snapdragon 5G Modem-RF Systems for Apple smartphone launches in 2024, 2025, and 2026.5 Apple, Samsung, and Xiaomi each represented 10 percent or more of Qualcomm’s consolidated revenues in its FY2025 10-K, with explicit risk language about large customers building their own chips.6
The roadmap is also still moving. Qualcomm framed its X85 modem-RF as an eighth-generation 5G modem-to-antenna solution targeting smartphones, PCs, fixed wireless access, automotive, and XR, with AI-powered connectivity features.9 The X105 was introduced as a 3GPP Release 19-ready modem-RF system with up to 14.8 Gbps peak download, 4.2 Gbps peak upload, NR-NTN satellite support, a 6 nm RF transceiver, quad-band GNSS, and 5G Advanced features framed as a bridge toward 6G.10
Qualcomm still has the best complete engine. Apple is learning how to own the car.
VI. The real risk is unbundling
Qualcomm’s old pitch is structural. Buy the complete modem-RF system because the chain is too hard to integrate yourself. The pitch is real. Self-integration is genuinely difficult. Carrier certification alone has consumed years for new entrants.
Apple’s counter-strategy is also structural. Replace one layer at a time until Qualcomm no longer owns the system. Apple does not need to beat Qualcomm everywhere immediately. It only needs to demote Qualcomm from system owner to replaceable supplier.
The timeline tells the story.
VII. Why this matters beyond iPhone
The modem-RF stack is expanding outside phones. Android flagships, PCs, XR headsets, automotive cockpits and ADAS systems, fixed wireless access, industrial IoT, satellite connectivity, 5G Advanced devices, and early 6G testbeds all need credible cellular silicon. Qualcomm is trying to scale modem-RF leadership across these markets. Apple only needs to solve its own devices.
That difference is everything. Qualcomm is optimising for broad merchant platform coverage. Apple is optimising for its own products, battery life, industrial design, and ecosystem control. The two strategies produce different roadmaps, different RF tradeoffs, and different definitions of “good enough” on each tier.
Qualcomm builds for the industry. Apple builds for the iPhone.
Quick terms
- Modem / baseband
- The chip that processes cellular communication protocols.
- RF transceiver
- Converts signals between baseband and radio frequencies.
- RF front-end
- Amplifiers, filters, switches, and tuning components between modem and antenna.
- mmWave
- High-frequency 5G spectrum with very high bandwidth but short range.
- Sub-6 GHz
- Lower-frequency 5G spectrum used more broadly worldwide.
- MIMO
- Multiple antennas working together to improve throughput and reliability.
- Antenna module
- Integrated antenna hardware that transmits and receives radio signals.
- Carrier aggregation
- Combining multiple bands for higher throughput.
- NR-NTN
- 5G non-terrestrial network support for satellite-related connectivity.
- Modem-RF system
- The integrated chain from modem through RF and antenna components.
- 5G Advanced
- The later phase of 5G standards bridging toward 6G.
VIII. What could break the thesis
A serious piece needs counterarguments. The case for Apple unbundling Qualcomm has more than one honest failure mode.
- Apple’s modem could lag. Closing the gap to Qualcomm at the full system level can take longer than internal roadmaps assume.
- mmWave is genuinely hard. If Apple cannot deliver competitive mmWave, the US iPhone Pro tier may stay on Qualcomm longer than expected.
- Qualcomm keeps winning Pro iPhones. Apple’s 2023 extension already locked in modem supply for 2024 to 2026 launches.5
- Carrier certification edge cases. Global certification is unforgiving and small failures matter at scale.
- Qualcomm’s roadmap is fast. X85 and X105 are not stationary targets.910
- Apple may skip mmWave globally. If Apple judges mmWave deployment too uneven, it may permanently de-prioritise it outside select markets.
- Qualcomm has many markets. Android, automotive, PCs, XR, FWA, and AI devices can offset iPhone modem losses.
- Apple may still need Qualcomm. Higher-end features at the modem-RF system level may remain Qualcomm strongholds for years.
- Good enough is not enough at the Pro tier. A modem can pass tests and still be downgraded inside a Pro iPhone if RF performance lags.
- Vertical integration is expensive. Apple has the cash to absorb the cost. Most competitors do not, which limits how much industry copies the strategy.
The correct claim is not that Apple has already beaten Qualcomm. The correct claim is that Apple has started reducing Qualcomm’s system-level role inside the iPhone.
IX. Device-to-network control
The iPhone 12 antenna decision looked small in 2020. In hindsight it was a clue. Apple kept the Qualcomm modem, but separated one RF layer from Qualcomm’s complete system. Five years later, Apple shipped its own modem in the iPhone 16e. It started with sub-6 GHz because that is the mainstream global layer and the lower-risk entry point. mmWave remains harder. Qualcomm still leads in modem-RF integration.
The modem war is not just about who has the fastest chip. It is about who controls the full path from silicon to signal. Qualcomm still has the best modem-RF engine. Apple is trying to own the car.
1 Patel, D. (Oct 2020). Qualcomm Lost the iPhone 12 mmWave Antenna Module Contract to a Chinese Company. SemiAnalysis. Historical anchor for the USI mmWave antenna module decision and Apple’s willingness to separate layers of Qualcomm’s modem-RF stack. Used as inspiration only. No content, structure, or charts reproduced.
2 iFixit. iPhone 12 Teardown. Public teardown coverage of the iPhone 12 internals, including modem and RF-related observations. Used as context only, no images reproduced.
3 Apple. iPhone 16e Technical Specifications. Lists Apple C1 cellular modem and 5G sub-6 GHz with 4×4 MIMO; mmWave is not listed.
4 Reuters (Feb 19 2025). Apple reveals first custom modem chip, shifting away from Qualcomm. C1 framed as Apple’s first custom modem subsystem designed to reduce Qualcomm dependence, with reporting on process technologies and carrier testing scope.
5 Qualcomm (Sep 11 2023). Qualcomm Announces Agreement with Apple for Chip Supply. Snapdragon 5G Modem-RF Systems for Apple smartphone launches in 2024, 2025, and 2026.
6 Qualcomm Incorporated, FY2025 Annual Report on Form 10-K. Customer concentration commentary listing Apple, Samsung, and Xiaomi each as 10% or more of consolidated revenues, alongside risk language regarding large customers developing their own chips. Cited generically pending direct linking from Qualcomm investor relations.
7 Apple. iPhone 17 Technical Specifications (US). US iPhone 17 model supports 5G sub-6 GHz and mmWave, including bands n258, n260, and n261.
8 Apple. iPhone 17 Technical Specifications (Australia). Australian iPhone 17 model lists 5G sub-6 GHz; mmWave is not listed.
9 Qualcomm. Snapdragon X85 5G Modem-RF System product page. Framed as the eighth-generation 5G modem-to-antenna solution with AI-powered connectivity features targeting smartphones, PCs, FWA, automotive, and XR. Cited via Qualcomm product communications.
10 Qualcomm (Mar 2026). Qualcomm announces 5G Advanced leap with the Qualcomm X105 5G Modem-RF System. Release 19-ready modem-RF system with up to 14.8 Gbps peak download, 4.2 Gbps peak upload, NR-NTN satellite support, 6 nm RF transceiver, quad-band GNSS, and 5G Advanced framing.
11 GSA (Jul 2025). Millimetre Wave Spectrum, July 2025. 203 operators in 56 countries and territories investing in 5G mmWave; 24 operators in 17 countries with launched 5G networks using mmWave spectrum; additional operators deploying or testing mmWave.
12 Reuters / Bloomberg (Dec 6 2024). Apple plans three-year modem rollout to compete with Qualcomm. Multi-year Apple modem rollout, first generation in a lower-end iPhone, targeting reduced Qualcomm reliance over time, with the existing supply agreement through 2026 as context.
- Nvidia Didn’t Buy Arm. It Built the AI Factory Anyway. Companion essay on how Nvidia built the CPU-GPU-DPU stack without owning Arm.
- MediaTek and the Fragmented Compute War. Companion essay on a Taiwan fabless platform between Android, edge AI, automotive, and hyperscaler ASICs.
- The AI Memory Wall. DRAM, HBM, packaging, and semicap as the new center of computing.
- The Dry Resist War. Patterning as a strategic process technology for AI-era chipmaking.
- Qualcomm Investor Relations. Quarterly results, modem-RF roadmap, customer concentration disclosures.
- Apple Newsroom. iPhone announcements and regional spec sheets.
- The AI Field Manual. Reference layer for the AI stack: hardware, memory, models, agents, safety, economics.
This is Essay No. 016. The topics: intelligence, AI, systems, knowledge, and the questions underneath the questions everyone else is asking. If you read this far and disagreed with any part of it, write to me. I read everything.