Data Center
Scalable acceleration for the AI era
The agentic era requires a fundamentally different data center architecture. Qualcomm Dragonfly™ products redefine data centers with industry-leading performance and power efficiency across CPU, AI accelerators and connectivity, together with custom silicon, enabling businesses to achieve sustained and scalable AI infrastructure growth.
Industry-leading efficiency
at rack scale
Memory bandwidth and energy are the AI inference infrastructure bottlenecks. Qualcomm Dragonfly™ solutions are engineered for leading bandwidth, tokens-per-watt, and tokens-per-dollar.
Decode performance
XX
Tokens/second/watt decode sustained rate at agentic workload
Power efficiency
+XX%
Better tokens/watt vs GPU baseline at the same throughput
Cost reduction
-XX%
Lower TCO per million tokens vs. GPU-based inference stack
Throughput gain
+XX%
Higher tokens/second vs GPU baseline at same power
Faster first token
XX%
Lower prefill latency, lower time-to-first-token vs X86 CPUs
Bandwidth advantage
XXx
Higher memory bandwidth/watt compared to HBM4E-based solutions
Decode performance
XX
Tokens/second/watt decode sustained rate at agentic workload
Throughput gain
+XX%
Higher tokens/second vs GPU baseline at same power
Faster first token
XX%
Lower prefill latency, lower time-to-first-token vs X86 CPUs
Power efficiency
3-5x
Better tokens/watt1
Cost reduction
-XX%
Lower TCO per million tokens vs. GPU-based inference stack
Bandwidth advantage
XX
Higher memory bandwidth/watt compared to HBM4E-based solutions
1 vs contemporary GPU-based architectures on decode performance
Tokens / (second * watt)
up to8x
better than GPU based systems1
Memory bandwidth / watt
6x
higher than HBM based systems2
Memory capacity / watt
200x
better than SRAM based systems3
Performance / watt
>2x
better than the latest server CPU systems4
Effective memory bandwitdh
54x
higher per card (AI300 with HBC Gen 2 vs AI200)
Explore our offerings
CPU
Purpose-built server CPU for leading power efficiency, core density, and utilization.
AI Accelerators
Rack-scale performance at leading tokens-per-watt and tokens-per-TCO.
Connectivity
Rack-scale connectivity with proven optical and electrical DSPs.
Custom Silicon
Tailored custom silicon solutions with cutting-edge IP to meet the specialized demand for next-gen data centers.
CPU
Performance redefined. Efficiency reimagined.
The Qualcomm Dragonfly™ C1000 CPU is purpose-built for data centers and engineered for best-in-class power efficiency, performance elasticity, and rack density for agentic AI, general-purpose, and AI head node workloads. By optimizing for power, it enables high rack density and superior total cost of ownership.
250+
core count chiplet design
>2x
better performance/watt4
AI Accelerators
A generational leap in efficiency
Built for disaggregated AI inferencing, Qualcomm Dragonfly™ AI250 rack with Qualcomm® High Bandwidth Compute (HBC) Gen 1 provides 18x effective memory bandwidth compared to AI200, resulting in superior power efficiency and performance at industry-leading tokens-per-dollar.
Industry-leading
43TB
of memory per rack5
Industry-leading
768GB
of memory per card5
Over
7.4PB/s
effective memory bandwidth per rack with HBC Gen 1
Efficient
140kW
rack-level power consumption
Industry-leading
43TB
of memory per rack5
Over
7.4PB/s
effective memory bandwidth per rack with HBC Gen 1
Industry-leading
768GB
of memory per card5
Efficient
140kW
rack-level power consumption
A multi-generational roadmap. An annual cadence.
Qualcomm Dragonfly AI200
Rack‑scale AI inference platform focused on memory capacity and TCO.
Qualcomm Dragonfly AI250
A generational leap in efficiency and performance with HBC Gen 1.
Qualcomm Dragonfly AI300
3rd generation rack-scale AI inference platform, with increased bandwidth and deployment options.
Connectivity
Data movement at inference scale
High‑speed connectivity for the AI era, powered by advanced DSP technology optimized for PAM4 and Coherent‑lite. Built to drive 800G and 1.6T performance across optical modules, AOCs, and AECs.
Performance & scale
High-performance retiming delivers rock-solid 800G/1.6T connectivity with uncompromised signal integrity across AI and cloud fabrics.
Advanced DSP technology
Advanced DSP algorithms engineered to deliver best-in-class bit error rate (BER) performance with low power and low latency.
Diagnostic tools
A powerful suite of diagnostic and monitoring tools designed to give complete visibility into link performance.
PAM4 and Coherent-lite DSPs for optical transceivers
PAM4 DSPs for active electrical cables
Custom Silicon
Domain-optimized data center solutions
Overcome traditional hardware limitations with end-to-end custom silicon design. Leveraging advanced packaging, modular architectures, and leading-edge technology platforms, our approach delivers purpose-built silicon precisely engineered to maximize the performance of agentic AI while meeting the power and performance demands of modern AI workloads.
Qualcomm Dragonfly AI Infrastructure Management Suite
Orchestrate and scale efficiency
The suite provides provisioning, monitoring, orchestration and fault handling across rack‑scale deployments. Together, hardware, connectivity and software form the foundation of a cohesive data center platform approach — one designed to scale with customers as AI workloads evolve.
Partner momentum and broad ecosystem support
More software tools
Qualcomm® AI Inference Suite
A comprehensive set of ready-to-use AI applications, agents, tools, and libraries for developing and deploying AI inference on premises or via cloud deployments.
Qualcomm® Cloud AI SDK
Optimize trained deep learning models for high-performance inference on Qualcomm Dragonfly AI accelerators.
Contact Sales
We have an extensive technology portfolio designed to enable products across industries. Our sales team is ready to provide you with tailored assistance.
1 With Qualcomm HBC; Qualcomm estimates compared to contemporary GPU-based architectures on decode performance for select models.
2 With Qualcomm HBC; Qualcomm estimates compared to competing published product specifications normalized at card-level.
3 With Qualcomm HBC; Qualcomm estimates compared to competing published product specifications normalized at rack-level.
4 With Qualcomm Dragonfly C1000 CPU; Qualcomm estimates compared to existing product benchmarks for server CPU competitive offerings based on specs.
5 Compared to competing published product specifications normalized at card- and rack-level
