Qualcomm is losing its biggest customer, and it knows it. For years, the chipmaker leaned heavily on selling modems to Apple. That cash cow is dying fast. To plug the hole, Qualcomm dangled a bold headline number alongside its Q3 financial results: it expects to pull in $15 billion in annual datacenter revenue by fiscal year 2029.
That sounds like a massive victory if you say it quickly. Going from zero to $15 billion in compute hardware over a few years is no small feat. But look closer at the board. Against the titans already occupying hyperscaler server racks, $15 billion makes Qualcomm a minor player. It isn't a threat to the reigning heavyweights. It's a defensive scramble to build a new revenue stream before the smartphone modem check bounces for good.
The $15 Billion Bet to Replace Apple
The immediate pressure comes directly from Cupertino. Apple has spent years working to design its own in-house modems. Qualcomm executives finally admitted to investors that the divorce is moving faster than expected, as detailed in reporting from The Register.
During the Q3 earnings call, management warned of an accelerated step-down in Apple revenues. Qualcomm's share of modems in the next iPhone generation will drop materially below its previous 20 percent estimate. The company now advises investors to expect less than $2 billion in total annual sales to Apple next year. Chief Financial Officer Akash Palkhiwala didn't sugarcoat the situation, telling analysts that the shift "obviously accelerates kind of the exit of Apple revenue out of our model."
That exit leaves a glaring hole on the balance sheet. In Q3, Qualcomm reported $9.95 billion in overall revenue—a 4 percent drop year-over-year—while net income fell 25 percent to $2 billion. Wall Street responded by knocking 5 percent off the stock in after-hours trading.
To offset the loss, Qualcomm is leaning on a broader diversification strategy. Management targets $40 billion in total annual non-handset sales by FY 2029. Automotive silicon is slated to bring in $10 billion, Internet of Things (IoT) hardware is tasked with $14 billion, and datacenters are expected to generate the remaining $15 billion. Considering Qualcomm logged $3.3 billion in non-smartphone sales in Q3, reaching $40 billion in four years demands flawless execution.
Why Apple Left Qualcomm Scrambling
For over a decade, baseband modem chips gave Qualcomm immense leverage in mobile hardware. Every iPhone sold meant a steady stream of high-margin silicon revenue and licensing fees. But relying on a single mega-client always carries existential risk.
Apple’s multi-year effort to build custom modem silicon was never a secret. What caught the market off guard was the speed of the transition. With Apple modem sales dropping under $2 billion next year, Qualcomm cannot afford a slow enterprise rollout.
The $15 billion datacenter projection isn't an opportunistic expansion into AI because the market looks fun. It is a mandatory replacement strategy. If non-handset divisions fail to hit their targets, Qualcomm's overall revenue profile will shrink permanently as Apple completes its in-house modem transition.
The Brutal Reality of Datacenter Math
Setting a $15 billion goal for datacenters looks ambitious until you bench it against the existing competition.
Consider where rival chipmakers stand right now. In fiscal year 2025, both AMD and Intel each pulled in over $16 billion from their respective datacenter divisions. Both companies are expanding those units rapidly. Then there is Nvidia, which sits on an entirely different plane. Nvidia is currently on track to generate over $250 billion in annual datacenter revenue, driven almost single-handedly by global AI infrastructure demand.
Put Qualcomm’s 2029 goal next to those figures. In four years, Qualcomm hopes to reach a revenue level that AMD and Intel already passed back in 2025. By 2029, Nvidia could easily be operating in an entirely different financial realm.
If Qualcomm hits its $15 billion target on schedule, it won't dominate the enterprise. It won't even be a top-tier peer. It will be a secondary supplier occupying the outer edges of cloud infrastructure. That isn't necessarily a failure, but it frames the company's ambition accurately.
The Dragonfly Gambit: Cheap Inferencing
So what is Qualcomm actually selling to cloud operators? It isn't trying to out-muscle Nvidia's monster training clusters. Instead, Qualcomm is taking aim at low-cost AI inferencing.
The company is pitching its "Dragonfly" datacenter platform as a cost-effective alternative for running trained AI models. The thesis is straightforward: training massive foundation models requires monstrous compute clusters and expensive GPUs, but serving everyday model predictions requires efficient, cheaper silicon. If hyperscalers want to keep operational costs under control, they need lower-power inferencing chips.
It’s a logical pitch on paper. But it comes with steep adoption hurdles. Enterprise customers and cloud providers have built their software workflows around Nvidia’s CUDA ecosystem and AMD’s ROCm tooling. Switching to Qualcomm’s architecture for inferencing means re-engineering software pipelines and taking on integration friction. Unless Qualcomm’s price-to-performance advantage is overwhelming, buyer inertia will be tough to break.
On top of that, competition in cheap inferencing isn't limited to traditional chip designers. Cloud giants like Amazon, Google, and Microsoft are building custom ASIC inferencing hardware in-house. Qualcomm isn't just competing with Nvidia; it's competing with its own potential cloud customers.
Wafer Inflation and the Mobile Rebound
While datacenter products slowly ramp up, Qualcomm still has to survive on handset revenue today. And that core business faces immediate macroeconomic pressure.
On the earnings call, Chief Executive Officer Cristiano Amon foreshadowed price hikes across Qualcomm's product line. He blamed rising wafer costs and component price inflation, though he argued that the price increases wouldn't choke off smartphone sales. Amon noted that even a double-digit price bump from Qualcomm is "small when you compare it to the order of magnitude of the memory bill of materials," pointing to high RAM costs as the main bottleneck for phone buyers.
Qualcomm is betting that its mobile revenue will catch a boost from a recovery in China. Amon expects Chinese smartphone sales to bounce back now that phone vendors have cleared out excess warehouse inventory. The catalyst? A fresh cycle of Android flagship phones packed with local agentic AI features.
Whether agentic AI features can trigger a massive smartphone upgrade cycle remains an open question. But Qualcomm desperately needs handset stability in China to fund its long-term datacenter buildout.
Can Qualcomm Really Win a Second Front?
Qualcomm’s pivot is a necessary survival move. Relying on Apple was always a temporary arrangement, and modem margins were bound to compress eventually.
Building a $15 billion datacenter business from scratch is a massive task. The company possesses world-class silicon engineering talent, and the Dragonfly platform targets a real industry headache around inferencing costs. But entering the server room means competing on foreign terrain against incumbents with entrenched software ecosystems and deep hyperscale relationships.
Even if Qualcomm hits every target by 2029, $15 billion in revenue will keep it in the enterprise secondary tier. That might be enough to replace lost iPhone checks, but it won't turn Qualcomm into a datacenter giant overnight.
