Qualcomm Dragonfly brings a robust platform of AI Compute accelerators, CPUs, breakthrough technologies, networking, & custom-silicon under one roof.
Enter The Dragon - Qualcomm's New Dragonfly Brand Is Primed For AI, Offering A Full-Stack Datacenter Portfolio, Encompassing Connectivity, CPU, and AI Accelerators
The Computex 2026 teaser of the Dragonfly brand is now being officially unveiled as a full-stack data center ecosystem, driving next-generation AI and general-purpose compute.
In our previous posts, we talked about the Dragonfly C1000 CPU and the Dragonfly HBC memory solution. Now, we will look at what the initiative has on offer beyond those two technologies, and to start, we first have the Qualcomm Dragonfly AI accelerators.
Why Qualcomm Dragonfly?
Dragonflies are among the most efficient fliers in nature, with an optimized structure that drives extraordinary energy efficiency. Their four wings—independent but highly coordinated—drive precision, agility and instant adaptation to changing conditions. They are built for sustained motion, maintaining performance over time without compromise.
That’s exactly how we think about data centers in the agentic era. With Qualcomm Dragonfly, we extended our DNA in high-performance, low-power compute across three product lines — connectivity, CPUs and AI accelerators — built for evolving AI infrastructure demands. AI workloads can run reliably across gigawatt-scale deployments. With a focus on superior performance and extraordinary energy efficiency, Qualcomm Dragonfly reduces latency, improves inference speed and enables better economic outcomes in modern data centers.
Qualcomm Dragonfly AI300 (Card and Rack)
The Qualcomm AI platform already saw the introduction of the AI200 and AI250 accelerators last year. The AI200 accelerators are being sampled now, and AI250 is on track for 2027. The AI250 AI accelerator will be the first to accommodate the HBC Gen1 solution, offering up to 43TB of LPDDR capacities in air and direct-liquid cooling designs. While the capacities remain the same as AI200 (LPDDR5X), with HBC, AI250 will offer an 18x boost in effective bandwidth and 5x the bandwidth per watt.
By 2028, Qualcomm is expected to sample its next-generation AI300 compute accelerator series. These will retain the air and direct-liquid cooling rack options, and the breakthrough will be the introduction of the second-generation HBC called HBC Gen2. This further raises the bar of what HBC can do with a 54x boost in effective bandwidth versus the AI200 and 8x bandwidth per watt versus HBM-based solutions.
For AI300, Qualcomm is also investing in scale-up architectures with the inclusion of UALink (Ultra Accelerator Link) and ESUN (Ethernet Scale-Up Networking), & Scale out with copper and optical infrastructure. The following are the full highlights:
- Third-generation, air- and direct-liquid-cooled rack-level AI inference platform – following the introduction of the AI200 and AI250 solutions last October
- AI300 integrates breakthrough Qualcomm HBC Gen 2 technology for compute acceleration with integrated memory and increased effective memory bandwidth, designed for disaggregated inference deployments (AI250 uses HBC Gen 1)
- Enables industry-leading memory capacity and effective bandwidth, enabling high-throughput, low-latency performance for large language & multimodal model (LLM, LMM) inference and agentic AI workloads
- Expecting 4x-8x better performance-per-watt compared to existing GPU-based architectures on memory bandwidth per watt per card
- Scale up with UALink (Ultra Accelerator Link) and ESUN (Ethernet for Scale-Up Networking); scale out with copper and optical
- Commercial sampling is expected in 2028
Qualcomm Dragonfly Connectivity Platform
On the connectivity front, Qualcomm aims to broaden its solution offering with die-to-die, copper, optical, and campus reach interconnects. The company will harness leading-edge process nodes to deliver 112 Gbps, 224 Gbps, and up to 448 Gbps SerDes through active electrical cables within & between the racks. We also see the mention of Co-Packaged Optics & Network Packaged Optics, which shows that Qualcomm is also keen on entering the silicon photonics race, which is already being set up between the likes of NVIDIA and AMD.
For Scale-Out, a new QAM16 Coherent-lite optical solution is being developed with a range of up to 20 kilometers, while PAM4 Optical SerDes enables optical speeds at up to 2km.
In the 2026-2027 timeline, Qualcomm's Dragonfly Connectivity lineup will see the addition of the O200 (1.6T Optical Modules/ACOs) and CU200 (1.6T AECs), while 2028 will see the introduction of next-generation CO1600 (1.6T-LR/3.2T-FR2), O400 (3.2T Optical Modules), and CU400 (3.2T AECs).
- Broad connectivity portfolio spanning die-to-die, copper, optical, and campus-reach interconnects for next-generation AI data centers
- Supports high-bandwidth 800G and 1.6T connectivity across optical, AOC, and AEC applications, from intra-data-center links to campus-reach deployments up to 20 km
- Combines Qualcomm Technologies’ SerDes, PAM4, coherent-lite DSP, signal integrity, and telemetry capabilities to support scalable, high-performance AI infrastructure
- Addresses data movement bottlenecks that are central to AI data center performance in increasingly distributed, disaggregated, and bandwidth-intensive infrastructure
Custom-Silicon Solutions Available To All
Qualcomm is also going big in the custom-silicon space. The company will offer performance-optimized silicon, end-to-end co-design capabilities, advanced packaging solutions, a proven IP stack, and full execution (including design and high-volume production) to its customers.
We've already seen reports about Qualcomm working with custom chip design with Chinese firms, & the official announcement means that the company is already talking to multiple firms for custom-silicon development.
- Performance-optimized silicon at scale for next-generation AI and cloud data center infrastructure
- Bespoke custom silicon for agentic AI and other specialized workloads
- End-to-end co-design capabilities across silicon, system, and software to address customer-specific performance, power, and integration requirements
- Advanced packaging and modular architectures designed to improve performance, power efficiency, and scalability
- Proven IP and streamlined design execution to support faster time-to-market and reduced execution risk
- Execution from design through high-volume manufacturing, supported by ecosystem and supply chain relationships
Qualcomm’s Dragonfly platform marks a bold and comprehensive entry into the AI data center space, delivering a true full-stack solution that brings together high-performance AI accelerators, next-generation connectivity, advanced memory, and custom silicon under one roof.
As the Dragonfly ecosystem matures through 2027–2028, it has the potential to become a powerful, cost-effective alternative for hyperscalers and enterprises seeking high-performance, power-efficient AI solutions. The Dragon has officially entered the arena.
About the author: A Software Engineer by training and a PC enthusiast by passion, Hassan Mujtaba serves as Wccftech's Senior Editor for hardware section. With years of experience in the industry, he specializes in deep-dive technical analysis of next-generation CPU and GPU architectures, motherboards, and cooling solutions. His work involves not only breaking news on upcoming technologies but also extensive hands-on reviews and benchmarking.
Follow Wccftech on Google to get more of our news coverage in your feeds.
