Qualcomm’s 35% Drop Offers AI Upside Backed By Strong Legacy Cash Flows

Qualcomm (QCOM) has fallen 35% from its May 2026 highs, and the market is pricing it like a company in decline following tough quarterly results and weakness in smartphones. Look a little further out, though, and the setup looks compelling. The near-term weakness is concentrated in Qualcomm’s handset business, while the company’s decades of experience building low-power, high-efficiency chips could position it for a very different opportunity: AI infrastructure.

QCT handset revenue dropped 20% last quarter as smartphone OEMs cut chipset purchases and worked down inventory amid memory supply constraints. Apple is the bigger structural issue. Qualcomm expects to supply modems for only around 20% of iPhones in 2026, and none at all by 2027, as Apple doubles down on its in-house modem chips. Together, these pressures dragged fiscal Q3 adjusted earnings down to $2.21, about 20% year over year, with Q4 guidance pointing to further softness.

However, both pressures fade with time. Once the Apple exit fully works through the numbers by fiscal 2027, the comparisons stabilize, and the earnings base resets on a slightly smaller but steadier core business. The memory crunch squeezing handset OEMs today could also ease as supply catches up, easing cost pressure on Qualcomm’s customers and, in turn, its own chip demand. Qualcomm’s earnings are projected to decline from about $10.50 per share in FY ’26 to an estimated fiscal 2027 earnings of $10.20 per share. However, even on these depressed earnings, the stock still trades at just about 16x forward earnings. The smartphone and licensing business, which generated close to $10 billion in operating free cash flow, will stabilize, while the upside could come from the AI business. (See How Qualcomm’s cash flows compare with peers)

The AI Opportunity

Over the last several years, much of the semiconductor industry’s growth has come from data centers, driven by artificial intelligence. While AI training has largely been dominated by GPUs, a growing share of AI spending is shifting toward inference, the process of running trained models in production. Inference could account for roughly two-thirds of AI compute by 2029, while agentic AI systems that execute tasks, interact with software, and make decisions autonomously could further increase demand for efficient inference hardware.

This shift plays to Qualcomm’s strengths.

The gap between AI compute demand and power supply is widening. A large AI data center can be built in 12 to 24 months, but securing high-capacity grid connections in key U.S. markets can take 36 to 84 months. More than 2,000 GW of generation and storage capacity remains in the U.S. interconnection queue. Power availability is increasingly becoming a constraint on AI infrastructure expansion.

Power efficiency has long been central to Qualcomm’s business. Smartphones operate under strict battery and thermal constraints, forcing the company to maximize performance per watt, the same metric that increasingly matters in AI data centers as power and cooling become major constraints.

The Nuvia acquisition gave Qualcomm its custom Oryon CPU architecture, now deployed in AI PCs and being extended into data-center CPUs. That puts Qualcomm in a position to apply its low-power computing expertise to a market where efficiency can directly translate into lower operating costs.

What Qualcomm Is Doing Today

Rather than competing directly for large AI training processors, Qualcomm is targeting inference. Its AI200 and AI250 systems are designed around power efficiency, memory capacity, and total cost of ownership rather than peak compute. The company has now added AI300, creating a multi-generation inference roadmap focused on large language models, multimodal workloads, and agentic AI.

Qualcomm has also secured a marquee customer. Its agreement with HUMAIN targets 200 MW of Qualcomm AI200 and AI250 infrastructure starting in 2026.

And Qualcomm is no longer pitching inference as a standalone opportunity. The company has unveiled the Dragonfly C1000, a 250-plus-core Oryon-based server CPU targeting agentic AI and general-purpose workloads. Meta has signed a multi-generation agreement to use Qualcomm’s CPUs in its next-generation server fleet, with the first C1000 expected to enter production in 2028.

Qualcomm is targeting more than $15 billion in annual data-center revenue by fiscal 2029, potentially creating a meaningful second growth engine alongside its core mobile business.

The completed $2.4 billion Alphawave acquisition further strengthens this strategy by giving Qualcomm high-speed connectivity, custom silicon, and chiplet capabilities, putting it increasingly in competition with the likes of Broadcom (AVGO) and Marvell (MRVL)  for custom AI infrastructure and data-center connectivity. See How interconnects drove Marvell almost 3x higher.

The Bottom Line

The handset business faces real headwinds, particularly from Apple, and earnings are likely to remain under pressure near term. But that weakness is increasingly reflected in Qualcomm’s valuation, while the company’s large cash-generating core gives investors time to wait for its next growth engine to develop.

Navigating the fast-growing yet volatile AI space requires balancing these high-conviction bets with a broader strategy anchored by mature cash generators. A smart portfolio helps you stay invested by limiting the impact of market shocks. While consistently beating the market is a challenge, the Trefis High Quality (HQ) Portfolio is designed to make it an achievable goal. The HQ strategy has consistently outperformed its market benchmark since inception, delivering cumulative returns of over 105 percent.