Qualcomm Closed Every Open Loop — and the Tape Is Still Selling AI Infrastructure

Five quarters of 'not yet,' settled in one afternoon

Qualcomm's calls for the past year have read like a list of things that had not happened yet: a Huawei licence, an Apple exit, a data center business with zero revenue, and a memory shortage that kept shrinking the handset market underneath it. On 4 November the company closed most of that list. “We closed fiscal '26 with many of the uncertainties that have been weighing on our business now behind us” — Unidentified Speaker · 2026-11-04, Cristiano Amon said — specifically the QTL renewals, the loss of Apple as a customer, and the drag from memory pricing on the size of the mobile industry.

The market's response was to keep selling. Qualcomm peaked at $251.02 on 29 May 2026 and has since drawn down 31%; over the last 90 sessions the stock is down 6.3%, and it is still 14.2% under its 25 September high of $201.97. A fresh report landing on a name already in a drawdown is a different animal: the question is not whether the story improved, but whether it improved fast enough to matter to a tape that has spent the autumn de-rating AI hardware.

The Huawei question finally has an answer

Go back to February. Asked about the Huawei licence, QTL president Alex Rogers was explicit: “Really no update on the Huawei discussions. The discussions are still underway.” — Alex Rogers, President, QTL (Qualcomm Technology Licensing) · 2026-02-04 Same answer in November 2025. That was the single largest unquantified risk in the licensing franchise. This quarter it is gone — Qualcomm completed long-term agreements with Apple, Huawei and Transsion, held the QTL revenue run rate unchanged, and told investors there are no major handset OEM renewals until fiscal 2030. That is a decade-long annuity re-signed, not a growth story, and it deserves to be priced as ballast rather than upside.

What is quietly striking is what fell out of the conversation. Apple product revenue was the company's third-largest keyword theme in the July quarter; this quarter "Apple" does not appear in the curated theme list at all. Apple is now an arithmetic assumption rather than a debate: QCT handset revenue declines mid-single-digit sequentially into December as the modem business ramps down, and handsets fall below half of total revenue in fiscal 2027. The company has spent years promising this — “the growth that we are targeting in the non-handset areas... will replace the entire Apple product revenue within the year” — Akash Palkhiwala, Chief Financial Officer · 2026-07-29 — and fiscal 2027 is when the promise gets marked to market.

Data center: first dollar, worst tape

Data center revenue begins with more than $650 million in the December quarter, on the way to $5 billion in fiscal 2027, a $7 billion exit run-rate, and $15 billion by 2029. Management frames it as a demand story: “we expect to be constrained by supply rather than demand for upsides beyond the $5 billion” — Unidentified Speaker · 2026-11-04. Two hyperscalers carry the bulk of that $5 billion at significant scale, with a third on an accelerator in advanced discussions — the 3rd customer that analysts pressed on twice. Inside the number sits connectivity revenue, including a second optical connectivity customer, and high bandwidth compute that Amon says hit 0.3 picojoules per bit in the lab against a 0.5 design target.

The tape is not cooperating. Over the last 30 days the global AI data center theme is a top decliner, printing a 21% negative price return across 56 negative tickers — with Qualcomm on that list — and data center revenues as a cluster is down 22.7% over the same window. The company is arriving at "revenue" exactly as the market stops paying for the theme. There is also the governance detail nobody likes: the Amazon relationship, with its potential $60 billion ten-year opportunity, carries an at-the-money warrant structure that Amon called "the exception," not the template.

The margin math is the actual argument

Here is the tension in plain terms. Management reiterates a 30% long-term QCT operating margin target, but guides QCT EBT at 20% to 22% for the December quarter — including a three-point hit from folding data center into the segment — while promising that double-digit price increases restore baseline gross margin to only the low end of the 48%–50% historical range, and that custom silicon dilutes gross margin from there. The same promise was made in July: “when we get through it, we expect to be consistent with the historical gross margin range we have.” — Akash Palkhiwala, Chief Financial Officer · 2026-07-29 The reported record is already wobbling: gross margin of 53.1% in the June quarter, down 2.5pp year over year, with free cash flow less stock compensation of −$335 million and effective net cash of −$3.6 billion, i.e. a net debt position. Funding a 26%-of-revenue R&D load while gross margin steps down is why a 21x price-to-earnings multiple looks less like a bargain than a verdict.

And the handset backdrop offers no slack. Fiscal 2026 units fell low-to-mid teens; fiscal 2027 is planned flat-to-down; the December quarter guides QCT handsets down mid-single-digit sequentially. Raising prices into that is bold. Amon's defence — that increases are landing on new platforms, so there is “less subject to pre-buying effect” — Unidentified Speaker · 2026-11-04 — is a reasonable answer to a reasonable skeptic, but the China Android "June was the bottom" call is now a claim the company has repeated for three quarters and needs to defend.

The option the market hasn't priced

Buried in the prepared remarks is the most company-unique item of the quarter: SpaceX selected Qualcomm to power Starlink Mobile's nationwide terrestrial 5G network using Dragon Wing infrastructure platforms, massive MIMO radios and Layer 1 software. Deployment starts in 2027; revenue is "more of a fiscal '28" item. It appears nowhere in the market's curated top themes, which is precisely the point — a zero-revenue, option-shaped win against the world's most closely watched private company. Compare that to the Agentic smartphones claim, which is real but not unique: agentic surfaces showed up across this reporting cycle in everything from Booking's "horizontal personal agents" to Shopify's "Personal Agent Protocol." Qualcomm's edge position — the first agentic phones shipping on Snapdragon, an 80-billion-parameter model running on Dragon Wing IQ10 — is a genuine device-side claim, but it is shared water, not a moat.

The watchlist for the next two quarters is narrow. First, whether QCT gross margin actually bottoms near 48% and turns in the second half of fiscal 2027 — the entire bull case rests on that inflection. Second, whether the December data center number beats $650 million rather than merely meets it, given management's supply-constrained framing and the July precedent that “It's good to have scale. It's good to have scale.” — Cristiano Amon, President and Chief Executive Officer · 2026-07-29 Third, whether the handset market stops being set by memory availability and starts being set by demand. Resolve all three and the 31% drawdown looks like an entry; resolve none and this is a licensing annuity trading at a cyclical multiple with an expensive hobby attached.