Qualcomm Investor Day: One CPU, One Memory Technology, and a $40 Billion Target
TechFlow Selected TechFlow Selected
Qualcomm Investor Day: One CPU, One Memory Technology, and a $40 Billion Target
Customers are secured, the product sample timeline is set, and the financial model is established. The upcoming quarterly financial reports will serve as the first test of Qualcomm’s roadmaps.
Navigating Web3 tides with focused insightsBy Boyang
Edited by Xu Qingyang
On June 25, local U.S. time, Qualcomm held its 2026 Investor Day in New York.
Qualcomm unveiled a comprehensive roadmap for AI infrastructure in data centers, launching the Dragonfly C1000 CPU, the AI300 inference accelerator, and High Bandwidth Computing (HBC) technology. It also announced multi-generation collaboration with Meta, deepened partnership with Hugging Face, and acquisition of AI software company Modular.
Qualcomm’s fiscal 2029 non-smartphone revenue target
Financially, Qualcomm raised its fiscal 2029 non-smartphone revenue target to $40 billion—nearly double its previous long-term goal. Within that, data center business revenue is expected to exceed $15 billion in fiscal 2029.
In after-hours trading, Qualcomm’s stock surged as much as 16%.
01 Data Center Revenue to Exceed $15 Billion
During the event, Qualcomm CFO Akash Palkhiwala forecast that Qualcomm’s data center business would generate “billions of dollars” in revenue in fiscal 2027—and over $15 billion annually by fiscal 2029.
Looking at Qualcomm’s overall revenue mix, non-smartphone revenue from QCT (the semiconductor division) is projected to reach $40 billion in fiscal 2029—up from the $22 billion long-term target set in fiscal 2024.
By fiscal 2029, smartphone-related revenue will account for only about one-third of QCT’s total revenue.
The remainder will be driven by several growth engines: $10 billion from automotive, and over $14 billion from IoT. The IoT segment includes industrial, networking, and robotics ($8 billion), and personal AI and computing ($6 billion).
Profit guidance was also revised upward.
Analysts’ average estimate for Qualcomm’s adjusted earnings per share (EPS) in fiscal 2029 stands at $15.26, while Qualcomm’s own target exceeds $18—a gap that directly triggered the post-market stock surge.
CEO Cristiano Amon attributed this growth to evolving AI usage patterns. He argued that AI is shifting from simple question-answering toward agent-based applications—models capable of autonomously executing multi-step tasks. Such workloads demand low-power computation, precisely where Qualcomm’s mobile chip expertise excels.
Amon added that AI computing is expanding into automobiles, everyday electronic devices, and robots—chip demand in these areas will continue to “open up.”
02 Dragonfly C1000 Debuts; Meta Becomes First Customer
The centerpiece hardware announcement was the Dragonfly C1000—a CPU purpose-built by Qualcomm for data centers.
Built on custom-designed Oryon cores and leveraging a chiplet architecture, the Dragonfly C1000 integrates over 250 cores and operates at frequencies exceeding 5 GHz. According to Qualcomm’s performance benchmarks, its performance-per-watt is more than twice that of competing server CPUs.
The Dragonfly C1000 supports PCIe Gen 7 and CXL interconnects, employs low-power memory technologies, and features built-in RAS capabilities—including ECC, fault isolation, and error recovery. Its thermal solution supports both air and liquid cooling, and its rack design complies with the OCP ORv3 standard.
Rack configurations featuring the Dragonfly C1000 were also disclosed: equipped with 43 TB of DRAM, sampling is expected in fiscal 2026.
Qualcomm has defined three specialized variants of the Dragonfly C1000:
First, an “agent CPU,” optimized for high-throughput agent orchestration and low-latency interactive AI tasks.
Second, a general-purpose CPU balancing two objectives: achieving optimal TCO (total cost of ownership) performance when running first-party workloads, and maximizing vCPU (virtual central processing unit) performance for third-party elastic use cases.
Third, an AI head-node CPU, designed to handle host processing with minimal overhead—freeing XPU resources to operate at full capacity during generative AI computation.
What truly lends weight to the Dragonfly C1000 is Meta’s endorsement.
Qualcomm announced a “multi-year, multi-generation” agreement with Meta, under which Meta will deploy the Dragonfly C1000 across its next-generation server clusters, with volume production scheduled for the second half of 2028. Subsequent generations of the CPU are also included in the scope of collaboration.
Qualcomm CFO Palkhiwala noted that through smartphone chips and other existing products, Qualcomm already works with virtually all hyperscalers—“This isn’t a new relationship.” This statement implies Meta is likely not the sole negotiating party, and additional customers may still be in discussion.
Addressing concerns about Qualcomm entering the data center market “too late,” CEO Amon responded: “When people ask whether it’s too late to enter data centers now, you should consider scale and execution capability, engineering capability, or operations and supply chain.”
His point is that Qualcomm’s large-scale systems engineering expertise—forged during the smartphone era—remains highly relevant in this market.
03 AI Accelerators Plus HBC Aim to Break the “Memory Wall”
Beyond CPUs, Qualcomm also updated its AI accelerator roadmap.
Following the previously launched AI200 and AI250, the AI300 inference accelerator debuted at this Investor Day—completing a trio of annual product iterations.
The core philosophy behind this platform is “decoupled, rack-scale AI inference.” Tony Pialis, EVP and General Manager of Qualcomm’s Data Center Business, explained that agent workloads require coordinated operation of CPUs, AI accelerators, and interconnect technologies—not reliance on a single chip. What Qualcomm is doing now is integrating compute, AI, memory, and connectivity into a unified rack-scale platform.
Within this platform, memory remains a critical bottleneck—and Qualcomm’s answer is High Bandwidth Computing (HBC).
HBC is a technology designed to break through the “memory wall”—a term describing the bandwidth bottleneck arising from data movement between processors and memory in AI computation. HBC adopts 3D-stacked silicon technology to tightly integrate compute units and memory, following a near-memory computing approach.
Qualcomm provided several metrics to illustrate HBC’s potential.
AI250 equipped with HBC Gen 1 delivers effective memory bandwidth of 133 TB/s per card—18× higher than AI200 using LPDDR5X. AI300, powered by HBC Gen 2, achieves a 54× bandwidth improvement over AI200.
Compared to today’s mainstream HBM (High Bandwidth Memory), HBC delivers 6× higher bandwidth at equivalent power consumption. Against SRAM (static random-access memory), HBC provides 200× greater capacity at equal power draw.
In short, HBC dramatically increases data throughput per watt—directly impacting data center TCO. Commercial samples of AI250 are expected mid-2027; AI300 commercial samples will follow in 2028.
Interconnect solutions—Qualcomm’s longstanding strength—were also featured. The company offers die-to-die, copper cable, optical fiber, and campus-scale interconnects supporting 800 Gbps and 1.6 Tbps speeds—covering distances from within-data-center to up to 20 kilometers.
More than 35 technology ecosystem partners have publicly endorsed this roadmap—including Supermicro, Lenovo, SK hynix, Micron, Samsung SDS, and Arista.
04 Acquisition of Modular, Partnership with Hugging Face
Beyond hardware, Qualcomm is making aggressive moves in software ecosystems.
First, the acquisition of AI software company Modular. The deal values Modular at approximately $3.9 billion in Qualcomm stock, with closing expected in the second half of 2026—subject to regulatory approvals.
Modular’s flagship offering is an open, AI-native software stack enabling models to run across diverse chip architectures—including CPUs, GPUs, NPUs, and custom ASICs—without requiring developers to rewrite code for each hardware platform. Co-founded by Chris Lattner and others, Modular’s platform is widely viewed in the industry as an open alternative to NVIDIA’s CUDA.
Commenting on the acquisition, Amon stated that as agents expand across data centers and the edge, the industry needs a more open and modern software foundation. Through this acquisition, Qualcomm aims to deliver genuine deployment flexibility for customers operating in heterogeneous compute environments.
Second, an expanded collaboration with Hugging Face. The partnership comprises three pillars:
* Migrating Hugging Face’s internal and developer workloads onto data centers powered by Qualcomm Dragonfly;
* Enabling direct loading of over three million open models available on the Hugging Face platform onto devices and data center racks powered by Qualcomm platforms—streamlining the path from experimentation to deployment;
* Developing “Hugging Face Agent,” a tool to orchestrate AI workloads across hybrid device-cloud environments—dynamically allocating tasks based on performance, cost, and latency requirements.
Clément Delangue, Co-Founder and CEO of Hugging Face, explained: “We’re empowering our 16 million developers to effortlessly run open models anywhere—from your handheld device to full-scale data center racks.”
A concrete element of this collaboration is Hugging Face’s provision of access to Hugging Face PRO—including premium storage, compute, and collaboration features—for customers deploying on Qualcomm-powered devices or cloud systems.
This step lowers the barrier for developers building applications with open models.
05 Automotive, Robotics, and China
Beyond the data center narrative, Qualcomm also shared updates on other businesses.
In automotive, “automotive design wins pipeline” has grown to $65 billion, and Qualcomm raised its fiscal 2029 revenue target to $10 billion. Underpinning this demand is continued penetration of ADAS and autonomous driving.
The IoT business has been further segmented: industrial, networking, and robotics separately targeted at $8 billion; personal AI and computing targeted at $6 billion. Qualcomm believes agents will trigger a new upgrade cycle for intelligent connected devices. The company estimates the combined market size for these businesses will reach $1.7 trillion by 2030.
Regarding the Chinese market, Amon offered a brief response during the event. While current U.S. government regulations restrict AI-related hardware exports to China, he indicated Qualcomm plans to offer data center chips specifically engineered to avoid triggering export controls. Though he did not elaborate on technical specifics, this statement confirms that opportunities in China remain active.
Overall, Qualcomm’s Investor Day delivered a comprehensive signal. From the C1000 to HBC, from the Modular acquisition to the Hugging Face partnership, from the $15 billion data center target to the $18 EPS goal—each milestone is verifiable. Customers are secured, product sampling timelines are set, and financial models are laid out.
The next few quarters’ earnings reports will serve as the first real-world test of Qualcomm’s roadmaps.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
