Apple's own modem will cut Qualcomm's Apple revenue in half in a single quarter — arriving faster than Qualcomm had modelled

Every figure in this article is drawn from Qualcomm's third-quarter fiscal 2026 results, published 29 July 2026, and its earnings call, or from CNBC as attributed. Qualcomm's fiscal year ends in September. Sections marked as analysis are identified as such.

There is no surprise in Apple building its own modem. Apple bought Intel's modem business in 2019 for exactly this purpose, Qualcomm has told investors for years to assume the business goes to zero, and both companies have negotiated their agreements on that basis.

The surprise is the timetable. Qualcomm told investors last month that revenue from Apple will fall roughly 50 per cent in a single quarter, as Apple's second-generation in-house modem — the C2 — reaches volume faster than Qualcomm had modelled, with a reduced Qualcomm allocation in the iPhone 18 Pro.

The quarter

Revenue for the June quarter was $9,947 million with net income of $2,002 million — a solid result that reflects the old business rather than the new one. The interesting lines are underneath.

Automotive revenue reached a record $1.59 billion, up 61 per cent. Qualcomm has raised the annualised run-rate it expects to exit this fiscal year with to about $7 billion. Internet-of-things revenue grew 9 per cent. Between them, these are the businesses that have to absorb the loss of the largest customer in mobile phones.

Automotive is the more advanced of the two and the easier to verify, because car programmes are announced years ahead. A chip designed into a 2028 model year is contracted now, built into an infotainment or driver-assistance platform, and generates revenue for as long as that vehicle stays in production. The revenue arrives late and then stays — the opposite profile to a phone component, which arrives immediately and can vanish at the next model refresh.

The company's stated arithmetic is precise. Non-handset chip revenue is forecast to grow more than 60 per cent in fiscal 2027, fully replacing the Apple product revenue of fiscal 2026. By fiscal 2029, Qualcomm expects non-handset products to be roughly two-thirds of chip revenue and about $40 billion.

The data-centre bet

The most ambitious part of the plan is the newest. Qualcomm has set targets of $5 billion of data-centre revenue by fiscal 2027 and $15 billion by fiscal 2029 — from a business that currently sells essentially nothing into data centres. It has completed the tape-out of its first High Bandwidth Compute chip, with silicon demonstrations promised over the coming quarters and the first customer solutions in mid-2027.

The strategic logic is that AI inference — running trained models rather than training them — is a power-constrained problem, and power efficiency is the one thing Qualcomm has spent thirty years optimising for in a device that has to fit in a pocket and not catch fire. Inference at scale rewards performance per watt over raw throughput, which is the argument every challenger to Nvidia now makes.

It is worth being sober about that argument. Everyone from AMD to a dozen startups has promised more inference performance per watt than Nvidia, and Nvidia's response has consistently been that the hardware is the easy part and the software ecosystem is the moat. Qualcomm has no equivalent to CUDA, and the customers it needs to win are the same hyperscalers currently designing their own accelerators in-house.

Analysis: the two questions this raises

The first is whether the replacement revenue is as good as what it replaces. Apple modem revenue was extraordinarily profitable: a single customer, one design win, enormous volume, no channel. Automotive revenue is won design by design, model by model, across dozens of manufacturers on multi-year cycles, and it comes with automotive-grade qualification costs and automotive-grade margins. Sixty-one per cent growth on $1.59 billion is genuinely impressive and it is not the same business.

The second is the timing gap. Qualcomm says non-handset growth replaces Apple's product revenue in fiscal 2027. The 50 per cent cut lands now. Even on management's own numbers there is a period in which the old revenue has gone and the new revenue has not fully arrived, and the data-centre business — the largest single component of the 2029 ambition — does not ship a customer solution until the middle of next year.

That is the honest shape of the risk: not that the strategy is wrong, but that the two curves cross later than the market would like, and Apple has just moved one of them to the left.

Qualcomm has the balance sheet to absorb a gap of a year or two — it is profitable, it generates cash, and it is not carrying the kind of debt that forces decisions. What it cannot control is how patient its shareholders are during the crossing, and transition stories in semiconductors have historically been valued at a discount until the new revenue is actually in the accounts.

Analysis: what Qualcomm still owns

It is easy to overstate the damage. Qualcomm's licensing business — the patent royalties collected on essentially every 3G, 4G and 5G handset sold anywhere, regardless of whose chip is inside — is unaffected by Apple designing its own modem. Apple still pays to use the standards. That business is smaller than the chip division and carries margins the chip division can only aspire to.

Qualcomm also remains the default supplier for the entire Android premium tier, and the arrival of on-device AI has, if anything, strengthened that position: running models locally is precisely the kind of power-constrained silicon problem the company is built around, and it is the reason handset average selling prices have been rising rather than falling.

There is a quieter benefit in losing Apple as well, which no executive will say out loud. A single customer at that scale distorts everything — pricing, roadmaps, factory allocation, negotiating leverage — and Qualcomm has spent fifteen years in periodic litigation with this one. A company whose largest customer is worth 10 per cent of revenue rather than 20 is a less profitable company and a considerably more independent one.

The most useful way to read this quarter is as a test of a proposition Qualcomm has been making for ten years — that it is a systems company that happens to have been enormously successful in phones, rather than a phone-chip company. Automotive at a $7 billion run-rate says the proposition has substance. The data-centre targets say management believes it completely.

Apple has simply set the exam a year earlier than expected. Qualcomm's answer will be visible in the fiscal 2027 numbers, when the non-handset business either does grow 60 per cent or does not — a claim specific enough that it can be checked, which is more than most transition stories offer.

One caution for readers checking that claim later: fully replacing Apple's product revenue is not the same as replacing Apple's profit. Qualcomm has been careful to frame the 2027 target in revenue terms, and nothing in the disclosure commits the company to matching the margin. The gap between those two statements is where the next two years of this story will actually be decided.