Qualcomm once defined itself by dominance in smartphone processors. Now the company bets its future on artificial intelligence that stretches from pocket-sized devices to warehouse-scale data centers. Executives laid out the vision in detail at the firm’s Investor Day in late June. They set ambitious financial goals and unveiled a new family of products under the Dragonfly name. The moves signal a deliberate effort to reduce reliance on handsets while chasing the explosive demand for efficient AI computation.
Shares reacted with a sharp pop before settling. Investors weighed the promise of new revenue streams against the reality that many of the promised chips remain years from volume production. Yet the strategy builds on technical strengths Qualcomm has honed for years in low-power design and connectivity. And it comes at a moment when customers seek alternatives to power-hungry graphics processors for running AI models.
Cristiano Amon, Qualcomm’s president and chief executive, captured the shift. “We are defining Qualcomm’s next chapter as we accelerate our edge diversification strategy, introduce a comprehensive roadmap for next-generation AI data centers, and evolve into a platform company,” he said in the company’s official release. The statement underscores how the firm now positions itself across what it calls the full compute continuum.
That continuum runs from personal devices that perform on-device inference to industrial systems and hyperscale infrastructure. Qualcomm raised its fiscal 2029 non-handset revenue target for its QCT business to $40 billion. That figure nearly doubles the prior outlook. Within that total, data center AI infrastructure should exceed $15 billion. Automotive revenues are eyed at $10 billion. Internet of things and related segments add more than $14 billion, with industrial, networking and robotics contributing $8 billion. Personal AI and compute round out the picture at $6 billion. Handsets, long the core, would represent roughly one-third of the total.
The company also guided for non-GAAP earnings per share above $18 in fiscal 2029. Those numbers reflect confidence that agentic AI workloads will drive upgrade cycles across intelligent devices. Agents that act autonomously rather than simply respond to prompts require efficient local processing. Qualcomm believes its expertise in balancing performance and power consumption gives it an edge there.
Central to the data center push sits the Dragonfly portfolio. It includes the Dragonfly C1000 CPU, High Bandwidth Compute architecture, the Dragonfly AI300 inference accelerator and supporting connectivity silicon. The AI300 joins earlier designs in a multi-generation roadmap that aims for annual improvements in inference performance, energy efficiency and total cost of ownership. Qualcomm acquired Modular in a roughly $4 billion all-stock deal to bolster the software side. Modular’s technology lets AI models run across varied hardware without heavy rewriting. It also brings a programming approach that could challenge the lock-in some competitors enjoy.
Memory plays an outsized role in the architecture. The High Bandwidth Compute design promises eight times the tokens per watt and six times the bandwidth per watt compared with conventional high-bandwidth memory approaches, according to analysis in Forbes. That focus on efficiency matters when power consumption and cooling costs dominate data center economics. Custom silicon for hyperscalers will begin ramping in early fiscal 2027. Inference accelerators follow in the second half of that year. Server-class Oryon CPUs based on the architecture acquired with Nuvia arrive in mid-2028.
Early revenue in fiscal 2027 is projected near $5 billion, largely from custom silicon and connectivity products gained through the Alphawave acquisition. The $15 billion data center target for 2029 therefore depends on execution across multiple product generations that do not yet ship in volume. Some observers note the catch. “Qualcomm’s data center ambition is no longer aspirational,” the Forbes piece states, while highlighting that investors essentially underwrite a detailed roadmap rather than proven shipments.
Yet customer traction already appears. Meta signed a strategic multi-generation agreement to use Qualcomm’s data center CPUs. The deal covers two generations and positions the chips for agentic workloads inside the social media giant’s infrastructure. Partnerships with Hugging Face aim to advance open-source AI models that run efficiently from device to cloud. These alliances lend credibility. They also suggest hyperscalers want options that deliver strong performance per watt rather than raw peak flops.
The strategy echoes Qualcomm’s earlier success in automotive. The company first entered vehicles through connectivity and digital cockpits before expanding into advanced driver assistance systems. In data centers it follows a similar sequence: connectivity first, then custom silicon, inference accelerators and finally full server CPUs. That stepwise approach reduces risk. It also lets the firm leverage existing relationships with manufacturers who already buy its modem and radio frequency components.
On the edge, Qualcomm continues to push Snapdragon platforms that integrate powerful neural processing units. The latest designs deliver up to 80 trillion operations per second. Such capability supports always-on large language models and low-latency agentic features while preserving battery life and user privacy. Apple has highlighted on-device processing for future Siri enhancements. Automotive and industrial applications stand to benefit from similar localized intelligence that reduces dependence on constant cloud connectivity.
The total addressable market across AI data centers, edge computing, robotics, automotive and related areas could reach $1.7 trillion by 2030, according to Qualcomm’s estimates shared at the investor event. That figure captures the distribution of compute that executives believe will define the agentic era. Physical AI in robotics and industrial settings represents another growth vector. Here too the company sees opportunities to combine its low-power processors with connectivity know-how.
But challenges remain. Competition in data center AI stays fierce. Nvidia dominates training and holds strong positions in inference. AMD and Intel push their own accelerators and CPUs. Custom silicon from hyperscalers themselves could limit third-party opportunities. Qualcomm must prove that its efficiency advantages translate into meaningful wins at scale. Software maturity after the Modular deal will prove decisive. Developers need compelling reasons to adopt new tools rather than stick with established frameworks.
Valuation offers some cushion. As The Motley Fool observed in a June 28 analysis, the stock trades at a price-to-earnings ratio around 21. That sits well below the broader technology sector average near 44. The discount reflects skepticism about the new initiatives. It also leaves room for upside if the financial targets materialize.
Recent coverage reinforces the momentum. A June 26 piece on 24/7 Wall St. echoed the observation that the $15 billion target rests on silicon not yet in customers’ hands. Still, the article noted the 13 percent after-hours jump that followed the announcements before some retracement. CNBC reported on June 24 that the Dragonfly C1000 CPU and Meta deal helped drive positive sentiment.
Qualcomm’s bet goes beyond individual products. It aims to become a full-stack provider that supplies silicon, software, connectivity and reference designs. The approach mirrors how the company succeeded in mobile phones by offering integrated platforms rather than discrete components. Success in AI would require similar coordination across hardware and software teams plus close collaboration with cloud operators and device makers.
Executives pointed to multiple inflection points over the next three to five years. Agentic AI should spur device upgrades. Industrial and physical AI will demand new edge infrastructure. Six-generation wireless networks will integrate AI-native features. Each area plays to Qualcomm’s historical strengths in power efficiency, system integration and radio technology.
Whether the vision delivers depends on flawless execution. The Dragonfly roadmap must hit its cadence. The Modular integration needs to accelerate developer adoption. Customer wins beyond the announced Meta deal must follow. And power-efficiency claims must hold up under real-world large-model workloads.
Yet the company enters this phase with substantial resources and a track record of adapting to platform shifts. From 2G modems to 5G systems-on-chip to on-device AI engines, Qualcomm has repeatedly expanded its capabilities. The current push into data center AI represents the largest such evolution in years. If it succeeds, the firm could emerge as a diversified computing leader rather than a handset-centric supplier.
Analysts will watch the next several quarters for design wins, early revenue traction and progress against the 2027 milestones. For now the market grants Qualcomm the benefit of the doubt. The stock reaction, the detailed roadmap and the high-profile customer agreement all point to a credible attempt to remake the company for the AI age. Execution will decide if that attempt becomes lasting success.
Qualcomm’s High-Stakes AI Pivot: From Phone Chips to Data Center Powerhouse first appeared on Web and IT News.