By Zihao Recently, at its 2026 Investor Day, Qualcomm unveiled an ambitious goal: to achieve $40 billion in non-mobile revenue by fiscal year 2029, nearly doubling its previous guidance of $22 billion. Among these, the data center is defined as its largest incremental market in the future, with revenue expected to exceed $15 billion in FY2029. By then, the proportion of the mobile chip business in Qualcomm's total semiconductor division revenue will be compressed to one-third, marking Qualcomm's transformation from a "mobile phone chip supplier" to an integrated computing platform enterprise covering device, edge, and cloud scenarios.
However, such a massive business shift for Qualcomm is no easy task. What seems to ride the industry trend is in fact a gradual strategic game. Stepping out of the comfort zone and entering new fields requires not only determination but also solid strength.
01. Mobile Market Under Pressure, AI Data Centers Open Up a Second Growth Pole The smartphone sector has been Qualcomm's commercial foundation for decades. Its Snapdragon chips hold an absolutely dominant position in the Android flagship market, and the patent licensing business brings stable, high-margin cash flow. However, as the consumer electronics industry has evolved, smartphones have long bid farewell to incremental expansion and entered a cruel era of stock competition.
On the one hand, global smartphone shipments have long been maintained in the stock range of 1.2 billion to 1.3 billion units. Incremental growth in emerging markets has dried up, and the replacement cycle for users in mature markets has been extended to over 30 months. The overall industry growth rate has long remained in single digits or even negative growth, which cannot support Qualcomm's expectations for sustained high growth.
On the other hand, the continuous loss of market share has become a visible hidden danger. MediaTek continues to seize the Android market with high cost-performance chips, squeezing Qualcomm's shipment space; Apple's self-developed A-series chips have decoupled from Qualcomm's basebands, continuously compressing Qualcomm's high-end baseband orders; domestic mobile phone manufacturers are accelerating the promotion of self-developed chips, further diverting chip procurement demand. Under the superposition of multiple factors, the revenue growth rate of Qualcomm's mobile chips has significantly slowed down, and the weak risk resistance of a single business is an undeniable fact. Once the mobile phone industry enters a downward cycle, the company's overall performance will be directly under pressure. Finding a second growth curve has become an urgent priority.
In fact, Qualcomm realized the limitations of a single business very early and has continuously expanded into the automotive and IoT (Internet of Things) fields. Objectively speaking, Qualcomm has achieved great results in these two business lines. The order scale of automotive digital cockpits and vehicle intelligent driving chips has exceeded $65 billion, and the industrial IoT and consumer smart hardware segments have maintained double-digit growth for many consecutive years, continuously contributing to revenue. However, in terms of revenue volume and growth ceiling, automotive and IoT can never completely offset the cyclical decline of the mobile business. The sum of their revenue scales is still difficult to support Qualcomm's trillion-level long-term growth expectations, and the pace of transformation has long been in the awkward situation of being neither here nor there.
Just when Qualcomm was struggling to find a truly new increment that could rival the mobile business, generative AI set off a global computing infrastructure boom. Cloud inference and device-edge-cloud collaborative computing power ushered in explosive demand. The data center track has become the hottest application field in the chip industry. Qualcomm saw the potential of this track and began to build up strength, hoping to forge it into its key growth lever for the company to break through the growth dilemma and reshape its industry positioning.
02. Asymmetric Competition, Building Differentiated Advantages Currently, the data center market has long been a red ocean battlefield. NVIDIA firmly occupies the large model training market with its CUDA software ecosystem and HBM high-bandwidth memory GPUs. Intel and AMD rely on the x86 architecture to deeply cultivate traditional server CPUs, and cloud vendors' self-developed ASIC chips continue to squeeze the survival space of third-party chip manufacturers. Qualcomm did not choose to confront the head players head-on, but took a differentiated entry path of the "low-power inference exclusive track", relying on its energy efficiency underlying capabilities accumulated from decades of mobile chip development to open up its own cloud survival space.
Qualcomm previously officially released a new full-stack data center solution, forming a complete rack-level computing power system covering computing, AI inference, memory acceleration, interconnect transmission, and customized development. It is specially built for intelligent agent AI-intensive inference workloads, relying on decades of low-power chip R&D accumulation to build unique competitive advantages.
Qualcomm's underlying low-power architecture is its key technical moat. Currently, large global cloud data centers generally face the bottleneck of power supply limits. Data center expansion is no longer limited by chip procurement costs, but by power supply capacity, and the operating costs of high-power training GPU clusters remain high. In the decades of smartphone R&D, Qualcomm's Oryon CPU and Hexagon NPU have continuously refined energy efficiency optimization technologies, excelling in achieving high-density computing power output under extremely low power consumption constraints, which fits the most core pain point of the current data center industry—power costs and energy consumption constraints.
Cristiano Amon, President and CEO of Qualcomm, stated: "Agentic AI is driving a significant increase in data center AI inference demand. As agentic AI becomes a mainstream workload, infrastructure must deliver higher performance with lower power and lower cost. This aligns perfectly with Qualcomm's technological strengths, and we are well-prepared for this shift. With Qualcomm Dragonfly, we are bringing high-performance, low-power computing to the data center market."
As Qualcomm's flagship server CPU specially built for cloud scenarios, the newly launched Dragonfly C1000 server CPU is developed based on customized Oryon CPU cores, optimizing high-frequency performance with clock speeds breaking through 5GHz. It adopts a multi-chiplet interconnect architecture with over 250 cores, balancing high throughput, scalability, and single-core performance, adapting to three core workloads: intelligent agent scheduling, general computing, and AI management nodes. In terms of core energy efficiency, the performance per watt of this product is more than 2 times higher than existing mainstream server CPU competitors, making it the most energy-efficient and TCO-optimized computing chip in its class. This chip is expected to be officially commercialized in 2028 and will become Qualcomm's core lever to pry open the cloud general computing market.
The newly developed HBC high-bandwidth near-memory computing technology is another major highlight. It adopts an innovative dedicated near-memory computing architecture, integrating computing and ultra-high bandwidth memory through a 3D stacked silicon-based solution to solve the data movement bottleneck in AI computing. The AI250 equipped with the first-generation HBC technology can achieve an industry-leading 133TB/s bandwidth per card. Compared with the AI200 using LPDDR5X, the effective memory bandwidth is increased by 18 times; the AI300 equipped with the second-generation HBC technology has an effective memory bandwidth increased by 54 times compared to the AI200. Compared with the published board-level standardized product parameters of competitors, the bandwidth per watt supported by HBC technology is 6 times higher than that of HBM technology.
The Qualcomm Dragonfly AI300 accelerator, as the latest flagship of Qualcomm's AI inference product line, focuses on ultra-high effective memory bandwidth and ultra-large memory capacity, adapting to high-concurrency, low-latency inference scenarios such as large language models, multimodal large models, and agentic AI. In terms of energy efficiency performance, the AI300 has achieved a qualitative leap compared to traditional GPU architectures, with per-card memory bandwidth per watt increased by 4 to 8 times, matching the core needs of agentic AI for continuous inference, real-time response, and scalable expansion.
At the connectivity level, Qualcomm has created a high-speed connectivity product matrix covering all scenarios, including Die-to-Die chip interconnect, copper/optical cable high-speed connections, 800G/1.6T optical interconnects, and campus-level long-distance interconnects. It integrates core technologies such as self-developed SerDes, PAM4 modulation, and lightweight coherent DSP, supporting campus-level deployments up to 20 kilometers away, solving the data transmission bottleneck of distributed AI data centers, and completing the last link of computing cluster collaboration.
Qualcomm is also actively acquiring to make up for the shortcomings of the software ecosystem. It previously announced the acquisition of AI infrastructure software company Modular for $3.9 billion to make up for the full-stack software capabilities of compilers, development frameworks, and adaptation toolchains, solving the adaptation challenges of heterogeneous architectures. At the commercialization level, Qualcomm has already secured long-term customized orders from two global top cloud vendors, Meta and Microsoft. The Dragonfly C1000 is planned to be mass-produced in 2028 and deployed on a large scale in Meta's server clusters. At the same time, it has won customized silicon chip projects from multiple hyperscale cloud service providers, locking in medium- and long-term basic revenue in advance.
03. Risks and Uncertainties Still Exist Although Qualcomm is full of confidence and well-prepared, the practical problems to be faced still exist.
First, the construction of the server chip software ecosystem requires time accumulation, and it is difficult to shake the barriers of mature players in the short term. Even if basic software tools are made up through acquisitions, CUDA has accumulated a massive number of developers, models, and applications over more than ten years, and the migration cost for industry users is extremely high. Enterprise customers need to reconstruct models, debug algorithms, and adapt business systems when replacing underlying computing chips. Small and medium cloud vendors and AI enterprises generally tend to maintain the original NVIDIA ecosystem. Qualcomm needs long-term investment in developer support and ecosystem adaptation, making large-scale customer adoption difficult to achieve in the short term.
Second, the customer validation cycle for server chips is long, and revenue realization has a lag. The iteration cycle of mobile chips is about one year, and the terminal manufacturers' validation process is short; while data center server chips need to undergo years of stability, load, heat dissipation, and energy consumption full-scenario tests by cloud vendors. The interval from sample delivery to batch procurement is extremely long. Although Meta and Microsoft orders have been locked in now, the Dragonfly series will not be mass-produced until 2028, and large-scale revenue release will be in the 2028-2029 fiscal years. If the product performance and energy efficiency fall short of customer expectations, there is a risk of a significant downward adjustment of revenue targets.
Moreover, competition in the inference track continues to intensify, and track congestion compresses profit margins. The AI inference market locked by Qualcomm has become a common direction for the whole industry. AMD, Intel, Google, Amazon, and many domestic semiconductor enterprises are all laying out the inference computing track. Market supply continues to expand, and it is highly likely to fall into a price war in the future, compressing the profit margin of a single chip. The differentiated advantage established by Qualcomm relying on low power will be continuously caught up by competitors.
In addition, there are cyclical fluctuations in computing power market demand. The current AI infrastructure boom has driven explosive growth in computing power procurement, but once industry capital investment cools down and cloud vendors reduce server procurement budgets, data center business revenue will be directly under pressure. Whether the $40 billion non-mobile revenue target can be completed as scheduled highly depends on the continued high prosperity of the global AI industry in the next three years, which is destined to be a long-term tug of war.
04. Multi-point Breakthroughs, Collaborative Operations Many people say that Qualcomm is undergoing a "de-mobile" transformation, but Qualcomm has not abandoned its mobile business. Instead, it has built a strategy of technology interconnection and multi-track collaborative cultivation. The mobile business remains the basic chassis of the company's revenue and technological R&D.
In the stock competition mobile phone market, Qualcomm chooses AI as a new product differentiation breakthrough, breaking through the device-cloud technology barriers, and reversely reusing the large model inference, high-parallel computing, and low-bit quantization optimization technologies developed for the data center Dragonfly and AI300 cloud chips to the new generation of Snapdragon flagship platforms.
Currently, on-device generative AI has become the core competition point in the high-end smartphone market. Local operation of large models, intelligent agent interaction, and multimodal on-device processing put forward higher requirements for mobile phone NPU computing power and energy efficiency. The large-scale model compression and efficient inference scheduling architecture accumulated from cloud R&D can directly improve the AI processing capability of Snapdragon mobile terminals, realizing the smooth local operation of lightweight large models on mobile phones, thereby consolidating Qualcomm's monopoly position in Android high-end flagship chips and offsetting the pressure of share loss brought by MediaTek and terminal self-developed chips. At the same time, the mobile business continues to steadily contribute cash flow and patent licensing revenue, and the R&D experience of the underlying technical architecture continues to be output to automotive, IoT, and data centers, forming a technology sharing system for all business lines, making the stock mobile business a technology incubator and financial safety cushion for diversified transformation.
Automotive and IoT are the foundation of Qualcomm's diversified layout. In the vehicle track, Qualcomm continues to upgrade the Snapdragon Digital Chassis integrated solution, covering full scenarios of digital cockpits, vehicle intelligent driving, and vehicle-to-everything (V2X) communication. The revenue target for FY2029 has been raised to $10 billion. Unlike the short-term explosive computing track, the intelligent vehicle industry is in a ten-year long-cycle upward channel. The model iteration cycle is long, customer stickiness is high, and orders have extremely strong predictability, which can provide long-term stable and sustainable revenue.
The IoT track is divided by Qualcomm into two sub-segments: industrial network robotics and personal AI computing. The overall revenue target for FY2029 will exceed $14 billion, covering massive scenarios such as AI PCs, XR devices, industrial controllers, smart homes, and wearable devices. Among them, the Snapdragon X series AI PC chips have quickly opened the Microsoft ecosystem market, relying on low-power advantages to seize the thin and light AI notebook track; industrial IoT chips deeply empower intelligent manufacturing and edge gateway devices, seizing the dividend of industrial digital transformation.
From relying on Snapdragon chips to dominate the smartphone era, to actively breaking the dependence on a single track and laying out a device-edge-cloud full-domain computing ecosystem, Qualcomm is undergoing the most disruptive corporate positioning reshaping since its establishment. This is not only a self-reform for the enterprise to break through the growth bottleneck and reconstruct the valuation system, but also an active breakthrough to adapt to the AI computing power industry transformation and embrace the new future of the industry, and more importantly, a microcosm of major enterprises continuously seeking change in the AI era.