Qualcomm Unveils $15 Billion Data Center Revenue Target by FY29, Doubles Non-Handset Revenue Forecast to $40 Billion
Investor Day, June 24, 2026
Qualcomm delivered perhaps its most ambitious investor presentation in company history, revealing a dramatic expansion into data center infrastructure that management believes will generate $15 billion in revenue by fiscal 2029, with $5 billion arriving as soon as fiscal 2027. The company nearly doubled its non-handset revenue target for FY29 to $40 billion from the $22 billion forecast provided just 18 months ago, driven by aggressive moves into data center compute, automotive silicon, industrial AI, and robotics.
CEO Cristiano Amon framed the announcement as marking Qualcomm's next chapter, coming exactly five years after he assumed the top role. The company that once earned its reputation as the most focused semiconductor player in mobile is now positioning itself as a computing leader across the entire compute continuum, from sub-2 milliwatts to 200 kilowatts, unified by what management called the industry's first truly open AI software platform through its acquisition of Modular.
Data Center Infrastructure Emerges from Stealth Mode with Hyperscaler Commitments
Tony Pialis, who joined Qualcomm six months ago from Alphawave to run the data center business, revealed that the company has been operating in what Amon described as "submarine strategy" for several years, quietly assembling assets and capabilities. Pialis introduced Dragonfly, Qualcomm's comprehensive data center infrastructure platform designed specifically for agentic AI workloads that generate 50x to 100x more inference requests than traditional queries.
The centerpiece technology is High Bandwidth Compute, or HBC, which Pialis called a breakthrough in solving the memory bottleneck that has constrained AI accelerator performance. Unlike traditional GPU architectures that constantly shuttle data between compute units and memory stacks across thousands of wires, HBC places the AI accelerator directly underneath DRAM stacks. The result is what Pialis claimed as 200x better capacity per watt than SRAM solutions for ultra-low latency workloads, and 6x better bandwidth per watt than HBM-based competitors for high-throughput applications.
Pialis characterized this as delivering performance advantages of SRAM with the density and memory capacity of HBM, eliminating the congestion inherent in competing architectures. The company plans to launch its first HBC product, the AI250, in mid-2027, followed by the AI300 in 2028 with integrated scale-up and scale-out networking fabrics.
What gave the forecasts credibility were video endorsements from Satya Nadella of Microsoft and Mark Zuckerberg of Meta. Nadella specifically highlighted HBC's "innovative architecture with high memory bandwidth and integrated compute that unlocks significant improvements in cost and performance" for Azure data centers. Zuckerberg announced a "multigenerational collaboration for Qualcomm to supply CPUs for our data centers and help power our next-generation server fleet" to support Meta's goal of delivering personal super intelligence.
Pialis stated that for fiscal 2027, at least two global hyperscalers will each contribute over $1 billion in revenue, providing meaningful diversification. Custom silicon will drive the largest portion of the $5 billion FY27 target, with AI accelerators ramping in the second half and CPU revenue beginning in the second half of fiscal 2028. Importantly, he noted that custom silicon gross margins will run slightly below Qualcomm's corporate average but remain accretive at the operating margin level.
For the longer term, management targets capturing greater than 5% share of the $1 trillion addressable data center market within five to seven years, implying revenue potential well beyond the $15 billion FY29 figure.
CPU Architecture Positions for Agentic Computing Workloads
Pialis introduced the C1000 family of data center processors, claiming they will run the industry's fastest cores at greater than 5 gigahertz, which he said represents more than 30% faster clock speeds than any competitor. The processors will scale to more than 250 cores for high-throughput workloads and deliver greater than 2 terabytes of I/O bandwidth leveraging Alphawave's PCIe technology. Significantly, the C1000 will incorporate LPDDR memory for what management described as the highest performance, lowest cost memory solution, along with server-class security features and native HBC attachment for AI acceleration.
Amon explained that the CPU attach rates are "soaring through the roof" as agentic AI transforms data center economics, with customers unable to find sufficient CPU capacity. The C1000 family will span three segments: agentic CPUs with HBC attachment, general-purpose CPUs for virtualized workloads, and AI head node CPUs for orchestrating traffic across disaggregated heterogeneous compute clusters, targeting a $200 billion addressable market that management said grows with every analyst report.
When asked about arriving late to the data center market, Amon was emphatic. "Never too late for Qualcomm," he stated, noting that the company can win in fast-moving markets with technology leadership. He pointed to Qualcomm's consumption of over 1 million leading node wafers annually, 75 chip tape-outs per year with over 30 on advanced transistors, and the ability to go directly from mask completion to production at scale, ramping completely new nodes to 100,000 wafers within about two quarters.
Modular Acquisition Anchors Open Software Strategy
The surprise announcement of Qualcomm's acquisition of Modular represents what Amon suggested could be "an Android moment" or even "a Linux moment" for AI infrastructure. Chris Lattner, Modular's co-founder and CEO, has a pedigree that includes building the compiler technology running in every smartphone regardless of vendor, creating the Swift programming language at Apple, and architecting Google's TPU AI platform software stack.
Modular has spent four and a half years developing what it calls AI's unified compute layer, a software platform enabling AI models to run on any hardware and designed from day one for heterogeneity. Lattner emphasized this as "the portable alternative to NVIDIA's software stack, designed from day 1 for every AI accelerator." The stack includes Mojo for high-performance programming, MAX for model serving, and Modular Cloud for distributed serving infrastructure, achieving up to 50% faster inference workloads on third-party hardware.
Tim Davis, Modular's co-founder and president, described the platform as turning "heterogeneous data center systems into multi-silicon AI token factories" where enterprises can use the best silicon for each workload without vendor lock-in. Lattner stated the goal of building an open developer platform that will "grow into a full operating system built natively distributed, natively accelerated and agentically native by design."
Amon acknowledged that not everyone will immediately understand what Qualcomm is attempting but expressed confidence that as AI distributes everywhere and the industry demands an open ecosystem, Qualcomm's approach of supporting all participants will prove prescient. The company also announced a strategic partnership with Hugging Face, whose 16 million developers will drive demand for Dragonfly silicon, with agentic model onboarding automating the setup, optimization and deployment of models across Qualcomm's entire product portfolio with zero manual integration work.
Automotive Business Pulled Forward Again, Now Targeting $10 Billion in FY29
Nakul Duggal, who runs automotive, industrial and robotics, revealed that Qualcomm is again accelerating its automotive revenue trajectory. The company now expects to reach $10 billion in automotive revenue in fiscal 2029, pulling forward the timeline by two more years after having already advanced it by two years in the previous investor day 18 months ago. Management stated Qualcomm is on track to become the largest automotive semiconductor supplier globally, supported by a $65 billion design win pipeline that has grown from $45 billion two years ago.
The automotive business has delivered 23 consecutive quarters of double-digit year-over-year growth and will exit fiscal 2026 at $6 billion in annualized revenue. Content value has increased 8x from third to fifth generation products as digital cockpit capabilities expand, sensor counts multiply for ADAS, and generative AI comes to vehicles. Qualcomm now counts over 70 automakers and more than 100 Tier 1 and Tier 2 suppliers as customers globally, with 415 new car models launched since 2021, representing two new models every week for five years.
Duggal described automotive as "the first example" of physical AI, with the mixed criticality fabric of Qualcomm's Gen 5 platform allowing customers to run cockpit and ADAS applications separately or together as they choose. The company is running 30 billion parameter models on cockpits today commercially while concurrently supporting L2 to L4 autonomous driving stacks. Stellantis recently selected not only Qualcomm's Ride Pilot ADAS stack but the entire Snapdragon digital chassis for deployment starting in 2028.
Growth vectors beyond the current trajectory include robotaxis, which management expects to scale by decade end, token acceleration using HBC Gen 2 attached to automotive SoCs beginning in the 2028 timeframe, and AI/ML use cases for powertrain, drivetrain and battery management. The company acquired EdgeImpulse to enable ML ops locally in vehicles using the Snapdragon NPU for on-device AI compute.
Industrial and Robotics Positioned as Trillion-Dollar Long-Term Opportunity
Duggal outlined how Qualcomm has rebuilt its approach to industrial and embedded markets over the past 18 months, moving from generic IoT to vertical-specific solutions under the Dragonwing brand. The operational technology plane across industries is being rearchitected as AI processing moves to the edge, creating what management characterized as a once-in-a-generation upgrade cycle across billions of endpoints.
The company has built connectivity chips, camera processors, commercial processors and industrial processors addressing 12 verticals across industrial, commercial and mobility categories. Qualcomm now serves 38,000 unique customers through more than 35 leading distributors, 45 global system integrators, and over 200 hardware and technology partners. Indirect revenue has increased 77% from fiscal 2024 to fiscal 2026.
Vision AI represents a major unlock, with Qualcomm deploying a complete stack from camera chips and edge AI boxes to video AI services across retail, small and medium businesses, smart cities and venues. The company has also focused intensely on developer accessibility, acquiring Arduino with its 33 million developers, EdgeImpulse for model training and tuning, and Foundries for industrial-grade Linux. In August, Qualcomm will launch the Dragonwing VENTUNO Q with 40 TOPS of AI, octacore processing, 12 camera support, and a built-in safety island, running full upstream Linux and available for purchase on Amazon.
Robotics is where Duggal said "embodied AI gets physical," with Qualcomm targeting at least a $1 trillion opportunity over the next decade. The company has developed what it calls a hierarchical compute architecture spanning System 2 for reasoning, System 1 for action planning, and System 0 for millisecond motor control and reflexes. The Dragonwing IQ10, IQ9 and IQ8 processors are already shipping in production, powering humanoids, quadrupeds, cognitive arms, autonomous mobile robots and drones.
Qualcomm is building the full stack including simulation platforms for training robots in virtual environments before real-world deployment, data pyramids combining real-world, synthetic and open source data, and foundation models trained with behavioral cloning, teleoperations and reinforcement learning. David Reger, founder and CEO of NEURA Robotics, appeared in a video endorsing the partnership and describing how Qualcomm provides more than just compute, enabling robots to see, hear, feel, think and react autonomously through the NEURAverse deployment platform.
Management forecasts industrial, networking and robotics will contribute $8 billion in fiscal 2029, with the personal AI and compute segment adding another $6 billion for a combined IoT total exceeding $14 billion, representing a 20% CAGR from fiscal 2025.
Agentic AI Redefines Mobile Edge Devices and Creates New Form Factors
Amon devoted significant attention to explaining how agentic AI and orchestrators are fundamentally changing device architecture. Unlike the traditional model where the smartphone sits at the center of a user's digital life, Amon argued that the agent is now at the center, with devices becoming endpoints for agents. This creates a profound shift because devices now serve two users, humans and agents, each with different workflows and performance requirements.
The device must support both human-speed interactions through apps and agent-speed operations where AI systems manipulate the device autonomously on behalf of users. Amon said the epicenter of new agentic use cases is currently China, where agents already navigate devices and the web to complete complex tasks. He noted that once consumers experience agentic computing, the six billion people with smartphones will expect these capabilities on the same device they carry, not on a separate computer.
Perception and sensing are also transforming devices, with audio, visual and contextual data becoming critical inputs. This is driving entirely new device categories, particularly glasses and wearables that position sensors close to eyes, ears and mouth. Amon said Qualcomm has 40 different designs today with major AI and model companies exploring new form factors, with glasses representing the one form factor certain to achieve scale. The use case is simple but powerful: "see what I see and hear what I hear."
Amazon's Panos Panay appeared in a video describing the fundamental shift in computing that AI enables, stating that Amazon and Qualcomm are partnering on Alexa experiences that work seamlessly in the home and on the go as "AI can seamlessly move with people throughout their daily lives." Microsoft's Satya Nadella confirmed continued collaboration on Project Solara for agent-first devices following the PC reinvention with Windows, calling it "fantastic to see the reception since we announced it together earlier this month."
The token economics are also driving architectural change. Agents and orchestrators are generating 40x more annual token demand between 2026 and 2030 according to projections Amon cited. He demonstrated hybrid AI in action, comparing two computers given identical complex research and web design prompts. One ran entirely in the cloud using Claude and OpenAI GPT-4, while the other used smart routing with some models local and others in cloud. Both achieved identical outcomes, but the hybrid approach consumed dramatically fewer cloud resources through what Amon called a mix of experts approach.
Google's Rick Osterloh appeared in a video highlighting the companies' "shared full stack vision" combining Gemini models and Android system-level intelligence with Snapdragon silicon to deliver "distributed intelligence" that balances processing between cloud and edge for experiences that are "private, instantaneous and personalized." Osterloh said Gemini Intelligence will elevate the Android ecosystem across mobile devices, automotive, wearables, XR glasses and the brand-new Google Books laptop effort.
6G Designed as AI-Native Infrastructure for Token Generation
Amon explained that 6G is being architected specifically for the AI era, with the goal of transforming everyone into "walking cameras" capable of streaming high-definition video uplink across cell sites. While 5G enabled streaming high-definition video to devices, 6G will enable the opposite, allowing devices to stream what users see to provide contextual information for agentic experiences.
More significantly, 6G infrastructure is no longer dedicated communications equipment but is becoming part of the distributed compute architecture. The network will not just transport bits but will also generate tokens, with every radio frequency treated as a radar using models trained on RF characteristics for sensing everything from drone detection to object movement. This creates critical infrastructure for AI models and necessitates significant compute at the cell site edge.
The architecture mirrors the distributed compute model Amon described for data centers, with large centralized data centers, regional core network data centers, edge data centers, cell sites, and devices all participating in distributed inference. Some mobile operators will sell token generation capacity like cloud service providers for AI companies, making 6G foundational to what Amon described as "sovereign AI workloads."
Financial Targets Imply Qualcomm Can Scale to $100 Billion Revenue Long Term
CFO Akash Palkhiwala walked through the financial implications of the strategies outlined, describing how Qualcomm's business mix will transform radically. In fiscal 2027, handsets will represent less than half of revenue for the first time. By fiscal 2029, handsets will account for just one-third of revenues as non-handset businesses reach $40 billion, nearly double the $22 billion forecast made 18 months earlier.
The revised targets assume Android handset revenue grows modestly at 5%, which Palkhiwala emphasized does not account for any improvement in the memory supply environment or meaningful uplift from agentic AI experiences. Either development would provide upside to the handset forecast. Licensing revenue is expected to remain stable and scale with global 4G and 5G unit growth.
Qualcomm expects to achieve the $40 billion non-handset target with data center contributing $15 billion, automotive reaching $10 billion, personal AI and compute delivering $6 billion, and industrial networking and robotics adding $8 billion. The four-year CAGR from fiscal 2025 to fiscal 2029 across non-handset businesses is 40%.
Operating margin targets remain unchanged at 30% for QCT in the long term and 70% for QTL, though the mix will shift as data center scales. Custom silicon gross margins will run slightly below corporate average but remain accretive at the operating margin. Palkhiwala highlighted that Qualcomm has managed operating expenses carefully during its diversification, growing OpEx at only 6% over the past five years while revenue doubled, bringing OpEx as a percentage of revenue from 31% to 23%. Management expects further improvement to 19% to 20% as revenue scales.
The company targets EPS greater than $18 in fiscal 2029, more than tripling from the $6 achieved five years earlier. Capital allocation priorities remain investing in technology leadership and diversification, pursuing strategic M&A as evidenced by 35 acquisitions over five years, and returning most free cash flow to shareholders while maintaining a strong balance sheet. Over the past decade, Qualcomm has retired 30% of outstanding shares and returned $40 billion to shareholders over the past five years.
Palkhiwala concluded by noting that growth drivers extend well beyond fiscal 2029, including continued data center expansion, robotics becoming one of the largest markets long term, industrial upgrade cycles, ADAS and autonomy in both silicon and software stacks, personal AI device proliferation, and 6G deployments. He stated that with these opportunities, "we have an opportunity with the things we outlined to scale our revenue to $100 billion" long term.
During the Q&A session, management confirmed it has secured capacity and memory supply to support the $5 billion fiscal 2027 data center revenue target, with Amon noting that Qualcomm's scale and supplier relationships provide capacity allocation advantages. Tony Pialis emphasized that discussions with data center customers have shifted from megawatts to gigawatts, and deploying full infrastructure at a few gigawatts scale can achieve the fiscal 2029 targets. The connectivity products acquired through Alphawave are already qualified at a leading hyperscaler and generating meaningful revenue in the current fiscal year, with the two large custom silicon hyperscaler wins beginning production in calendar Q4 2026, which is fiscal Q1 2027 for Qualcomm.