PARAS MADAN
All writing

Latest in AI

Qualcomm AI Data Center Outlook: $650 Million Quarter and a $7 Billion Run Rate

Qualcomm’s latest earnings put a nearer-term revenue schedule behind its AI infrastructure plans, with an important distinction between annual sales and an exit run rate.

Qualcomm expects more than $650 million in fiscal Q1 2027 data center revenue and a $7 billion annualized exit rate, according to its November 4 earnings presentation. Both are forecasts, not reported sales.

The important question is how quickly customer programs turn into recurring shipments. A large long-term market target becomes more useful when a company attaches a quarter to the first meaningful revenue contribution.

What Qualcomm actually forecast

The fiscal Q4 presentation gives several numbers that describe different things.

MetricQualcomm’s outlookHow to read it
Q1 fiscal 2027 data center revenueMore than $650 millionExpected revenue for one quarter
Full fiscal 2027 data center revenue$5 billionPreviously stated annual target, reiterated
Fiscal 2027 exit pace$7 billion annualized run rateYear-end revenue pace expressed as an annual figure
Fiscal 2029 data center revenueMore than $15 billionLonger-term Investor Day target

The $7 billion run rate is not a forecast of $7 billion in fiscal 2027 sales. A business can finish a year at a much higher quarterly pace than it maintained during the earlier quarters. Keeping those measures separate matters when judging how much revenue the expansion will actually contribute.

The company says data center revenue should rise each quarter through fiscal 2027. That provides something concrete to check as results arrive, beyond whether management repeats its longer-term targets.

The customer work behind the ramp

Qualcomm lists design wins in connectivity, custom silicon and AI accelerators, and CPUs. Named customers include Amazon in the first two categories and Meta for CPUs.

These categories matter because an AI data center needs more than accelerators. Moving data and running the surrounding compute also create spending opportunities. Qualcomm’s disclosed customer mix points to participation across that system, rather than a single chip being expected to carry the entire expansion.

A design win still needs to become production volume. The useful evidence from here will be actual quarterly revenue and the durability of those customer programs, rather than treating every announced collaboration as already-earned sales.

The handset business is still under pressure

In its earnings release, Qualcomm reported fiscal Q4 handset revenue of $5.137 billion, down 26% year over year. Full-year GAAP company revenue was $43.12 billion. Quarterly GAAP revenue was $10.322 billion, while the quarterly non-GAAP figure was $10.442 billion.

That context explains why the data center timetable matters. The company is building another source of revenue while its largest established product category is weakening. It also makes sloppy comparisons particularly misleading: a quarterly company total should not be compared with an annual data center target as though the periods match.

There are costs to this expansion too. Starting in fiscal 2027, Qualcomm will report Data Center within QCT. The presentation says that inclusion reduces the first-quarter QCT earnings-before-tax margin guidance by three percentage points, reflecting investment ahead of the revenue ramp.

What to watch next

For AI infrastructure buyers, this is a signal of another supplier putting a commercial schedule behind its plans. It is not evidence that Qualcomm has already displaced existing systems, or that the forecast establishes superior performance per dollar. Those questions need workload results and actual deployments.

For the business itself, the next meaningful milestone is simpler: deliver the first quarter, then show that each subsequent quarter supports the promised ramp. The distance between a design win and repeat revenue is where this AI infrastructure story will be decided.

I’m Paras. I work at the intersection of AI, engineering and distribution.

I help technical founders explain complex products and turn their launches into content that developers and decision-makers can use.

Book a 30-minute call.