QCOM (Qualcomm) Investment Analysis

Against the backdrop of the AI giants in the U.S. stock market reveling in valuations soaring by tens of times, Nvidia's Jensen Huang has surprisingly done something unexpected—he's started "promoting" his potential competitors.

On June 8, 2026, at an event in Seoul, Nvidia CEO Jensen Huang suddenly made a public statement:
"Qualcomm is doing an amazing job. Buy their stock."

He even joked wittily: "I spent the whole day selling other companies' stocks. That's good; we should be happy for others' success."

As the absolute ruler of the AI empire, why would the "AI Pope" publicly and prominently "promote" this competitor, labeled by the market as a "mobile chipmaker," when his own graphics cards are selling like hotcakes and in short supply? What exactly is Qualcomm, the king of the mobile chip era who seems somewhat left out in the AI frenzy, secretly plotting?

Can Qualcomm really steal the next AI era right from under Nvidia's nose?

---

📌 Act One: The "Old Empire" Left Out in the AI Frenzy

Compared to AI concept stocks that often double in value, if you only look at the fundamentals, Qualcomm's core business might seem somewhat "mediocre."

The mobile phone market is saturated, with shipments peaking; major client Apple is secretly planning to fully develop its own modem; and the competition in the PC space is fierce with strong players everywhere. At first glance, this core business seems completely unrelated to the recent hot narrative of "AI computing infrastructure (10,000-card clusters, optical modules)."

In the eyes of many ordinary investors, Qualcomm is still just a "has-been" company that mainly relies on selling smartphone chips and collecting patent licensing fees (QTL) to get by.

But if you look down on Qualcomm because of this, you're making a big mistake.

In the eyes of professional investment institutions, this seemingly "unremarkable, modest-growth" core business is actually Qualcomm's most solid "cash cow."

High-margin patent fees and stable chip shipments provide Qualcomm with over $12 billion in free cash flow (FCF) every year. While the entire market is worried about the high valuations and "high bubbles" of AI stocks, Qualcomm sits firmly with an extremely robust financial foundation. This "old empire" core business constitutes Qualcomm's high safety margin.

---

📌 Act Two: The Stealthy "Inference Narrative" and the Ambition of Investment Banks

The development of the AI industry is approaching a critical inflection point: The demand for computing power is shifting significantly from cloud-based "training" to edge and local "inference."

The authoritative research firm IDC points out that as much as 75% of data computation is now moving outside the cloud (i.e., to the edge and local devices). As large model applications enter the deep waters of commercialization, the high cost of cloud APIs and unavoidable latency are driving computing power to accelerate its shift to intelligent terminals.

Coincidentally, Gartner also predicts that the cost of large model inference will plummet in the future, a change largely attributable to the accelerated adoption of edge devices in specific inference use cases.

Training a model is a one-time event (dominated by Nvidia), but running the model for inference is infinite. This precisely gives Qualcomm an opportunity for a stealthy advance. Qualcomm is comprehensively entering the core territory of AI inference from three dimensions:

1. Client-side PC (The AI-PC Era): Snapdragon series chips are becoming the "computing center" for local Agent intelligence on Windows PCs.
2. Custom ASIC Chips: Providing underlying technical support for hyperscale cloud providers' self-developed, differentiated inference chips.
3. Data Center Commercial Accelerators (AI200 / AI250): Designed with liquid-cooled rack-level architecture, specifically for low-cost operation of large model inference, completely moving away from expensive and out-of-stock GPU computing clusters.

This is by no means Qualcomm's self-indulgent narrative. J.P. Morgan's latest North American Equity Research report released in June 2026 shows that Qualcomm's data center business is undergoing explosive revaluation, with its projected data center revenue:

• Reaching $3 billion in FY2027
• Soaring to $35 billion in FY2031

Qualcomm's AI narrative is no longer just a promise on paper; it's a solid growth curve of tens of billions of dollars written into investment bank models.

---

📌 Act Three: Reconstructing the Valuation Logic — Buying Qualcomm and Getting an "AI Option" for Free?

If you break down Qualcomm's valuation ledger again (SOTP, sum-of-the-parts valuation), you'll find an incredibly shocking contrast in cost-effectiveness.

Qualcomm's current stock price is around $204, with an enterprise value/free cash flow (EV/FCF) multiple of only 14.6x. Compared to Nvidia, Qualcomm is a full four times cheaper.

We can break down Qualcomm into two parts for calculation:

• Core Mature Main Business (Mobile Chips + Patent Licensing): Even if we apply a 52% discount compared to peer company MediaTek (MTK), giving this part of the business only a 5x revenue multiple.
• Automotive AI Business (Smart Cockpit + Autonomous Driving): Qualcomm is racing ahead in the automotive field with a 38% growth rate. Comparing it to peers like Mobileye, we assign it a revenue multiple of 5 to 7x.

Calculating these two items, the basic foundation of mobile + patents + automotive already solidly supports an intrinsic value of $180 - $200.

What does this mean?

It means that the current stock price hardly assigns any valuation to Qualcomm's data center AI business and client-side PC AI narrative. Its pricing is approximately $0.

In other words, buying Qualcomm at the current price is equivalent to buying its stable core business and high-growth automotive business, while the entire potentially explosive "AI inference option" is given to you for free by Qualcomm.

---

📌 Act Four: The Duel of Shield and Spear

To grab a share of the pie under Nvidia's watchful eye, Qualcomm needs both a defensive "shield" and an offensive "spear."

🛡️ Qualcomm's Shield (Moat of Existing Businesses)

Qualcomm's defensive cards are extremely solid. On the edge side, it possesses an unshakable advantage in low-power management and SoC (system-on-chip) integration barriers. Even more formidable is that Qualcomm has quietly secured $45 billion in design pipeline orders in the automotive AI field. Whether it's smart cockpits or autonomous driving, it has already passed mass-production validation for L2+ levels with mainstream automakers like BMW. The continuous annual inflow of patent fee cash flow ensures its annual R&D budget of nearly $9 billion, keeping it at the forefront of underlying technology.

🗡️ Qualcomm's Sharp Spear (Three Core Advantages on the AI Inference Side)

On the offensive front, the spear Qualcomm thrusts at Nvidia is highly targeted:

• Advantage One: Ultimate "Performance per Watt" (Energy Efficiency Ceiling)
  Qualcomm's accelerators originate from the Hexagon NPU architecture of mobile SoCs. At a time when data centers face "power shortages" and thermal bottlenecks, Qualcomm's ultra-low energy efficiency ratio becomes the perfect solution to energy consumption and heat dissipation disasters.
• Advantage Two: Low-Cost TCO Path with 768GB Ultra-Large Memory
  Qualcomm's AI200 avoids the constraints of expensive and severely out-of-stock HBM (High Bandwidth Memory), supporting up to 768GB of large-capacity, low-power memory per card. This means cloud providers can easily run offline and concurrent inference for ultra-large language models at extremely low hardware and software procurement costs.
• Advantage Three: "Interception Rights" at the Edge (Competing a Level Down Against Cloud and Base Stations)
  As the absolute dominant player in mobile/PC/vehicle infotainment systems, Qualcomm has a natural dominance at the edge. When lightweight Agent intelligence becomes widespread, Qualcomm can "intercept" inference tasks at the edge and run them locally:
  • Compared to Cloud Inference: Edge inference offers zero latency, extremely high data privacy, and the absolute advantage of eliminating cloud API fees.
  • Compared to AI-RAN (Base Station-side AI): AI-RAN still faces congestion and signal jitter in wireless channels, while edge inference latency is constant at the microsecond level. Furthermore, AI-RAN still requires base stations to bear extremely high power consumption, whereas edge inference distributes the computing power consumption across the batteries of billions of terminals. Most crucially, highly personalized local context data no longer needs cross-network synchronization but completes the loop directly at the source.

However, shadows also lie on this long spear. Qualcomm still faces two hidden reefs: first, the potential geopolitical policy red lines affecting ByteDance's custom inference chips in China (impacting nearly 30% of Qualcomm's revenue in China); second, cloud providers are already deeply dependent on Nvidia's CUDA software ecosystem, making migration costs extremely high.

---

📌 Act Five: The Big Showdown in New York

All the gamesmanship, doubts, and expectations point to an impending decisive moment: June 24th, the Qualcomm Investor Day in New York.

This is the most anticipated "reveal of the cards" in the semiconductor field for the second half of the year.

The entire industry and Wall Street are holding their breath: Can Qualcomm, on this day, pull out actual hard procurement contracts with hyperscale cloud providers like Meta and AWS on the spot, declaring its successful surprise attack on Nvidia?

Or, as market pessimists fear, will it only present a qualitative outlook with grand blueprints but lacking actual hard customers, leaving the suspense hanging in the air?

What is the subtext behind Jensen Huang's high-profile "promotion"? Can Qualcomm deliver an answer sheet that stuns the entire semiconductor market?

June 24th, New York. The cards are about to be revealed.

Reference Sources

1. Qualcomm Unveils AI200 and AI250 Official Press Release — Qualcomm's official announcement of the AI200/AI250 rack-scale data center inference accelerators; source for the technical details in the "Sharp Spear" section