AI Infra Summit 2026: Building infrastructure for the agentic AI era
What you should know:
- Qualcomm is bringing its Qualcomm Dragonfly portfolio to AI Infra Summit 2026, showcasing a rack-scale platform purpose-built for agentic AI inference.
- Tony Pialis, EVP & GM, Data Center, will share Qualcomm's vision for AI infrastructure, centered on a new metric for the AI era: tokens per watt, not FLOPS.
- Vinesh Sukumar, VP, Product Management, will highlight how Qualcomm is positioning itself at the forefront of the agentic AI era by championing a distributed AI architecture.
- Demonstrations will highlight breakthroughs in inference efficiency, memory architecture, connectivity, sovereign AI and enterprise-ready agentic AI deployments.
- A Hackathon is scheduled to bring together developers and innovators building on top of Qualcomm platforms.
As AI moves from generating responses to reasoning, planning and acting autonomously, infrastructure requirements are changing fundamentally. Agentic AI introduces a new class of workloads that demand dramatically higher inference capacity, greater memory efficiency, lower latency and more sustainable economics.
At AI Infra Summit 2026, Qualcomm Technologies is showcasing how it is helping the industry make that transition. The shift toward agentic AI plays directly to Qualcomm Technologies’ strengths. For decades, we have focused on solving complex compute challenges under strict power constraints. Today, we're bringing that expertise to the data center through Qualcomm Dragonfly, our rack-scale AI infrastructure portfolio designed for the next generation of inference workloads. Our vision is straightforward: the future of AI infrastructure will be defined by how efficiently systems generate useful AI work, making tokens per watt a more meaningful measure of performance than traditional compute metrics alone.
Tony Pialis: Reshaping infrastructure for the agentic era
During his AI Infra Summit keynote on Wednesday, Sept. 16 at 2:40 p.m. on the Main Stage, Tony Pialis, EVP & GM, Data Center, Qualcomm Technologies, Inc., will expand on the strategy we introduced at Investor Day earlier this year: agentic AI is driving the industry's transition from infrastructure optimized for training toward infrastructure optimized for large-scale inference.
As agentic workloads continue to scale, data centers face increasingly difficult challenges around power consumption, memory bottlenecks and total cost of ownership (TCO). We believe the industry must move beyond traditional approaches that prioritize peak performance at any cost and instead embrace architectures built around efficiency, scalability and system-level optimization.
At the center of this strategy is Qualcomm Dragonfly, a rack-scale platform spanning IP, silicon, cards, software and complete rack solutions across four product areas: CPUs, AI accelerators, custom silicon and connectivity. Together, these technologies are designed to help hyperscalers, cloud providers and enterprises deploy AI infrastructure that delivers more intelligence within the same power envelope.
Qualcomm Dragonfly solutions are set to drive industry-leading TCO through superb efficiency, with metrics including:
- Up to 8x better tokens per second per watt versus GPU-based systems1
- Up to 200x greater memory capacity per watt versus SRAM-based approaches2
- Up to 6x higher memory bandwidth per watt versus conventional HBM-based designs3
- More than 2x better performance per watt than the latest server CPU systems4
These innovations are designed to address one of the central challenges of modern AI infrastructure: keeping increasingly sophisticated models fed with the memory bandwidth and capacity they require while maintaining sustainable power consumption.
As Tony often notes, the next phase of AI won't be constrained by compute alone. It will be defined by how efficiently infrastructure can transform energy into useful AI output.
Connectivity matters more than ever
As AI factories grow larger and more distributed, networking becomes a critical determinant of overall system performance. During the summit, Tony will join Lisa Spelman, CEO of Cornelis Networks, during her session, “The Network Is the Computer — and It's Time It Acted Like One,” on Tuesday, Sept. 15 at 3:00 p.m. on the Data Movement Track, to discuss the evolving role of connectivity in large-scale AI systems.
The discussion will explore how efficient data movement, advanced fabrics, optical networking and tightly integrated connectivity architectures are becoming increasingly important as agentic AI workloads scale across racks, clusters and entire data centers.
Powering scale across the AI compute continuum
Qualcomm Technologies' presence at AI Infra Summit extends beyond the data center. On September 17 at 9:55 a.m. on the Main Stage, Vinesh Sukumar, Vice President of Product Management, will deliver a keynote, "Powering Scale: The Evolution of AI Systems," exploring how AI is evolving from conversational experiences into autonomous agents that reason, take action and operate across devices, networks, enterprises and cloud environments. As agentic AI drives exponential growth in token demand, Vinesh will discuss why the industry's next challenge is not simply building more compute, but building AI systems that can deliver lower cost, lower latency and greater efficiency at scale.
His presentation will highlight Qualcomm Technologies' vision for a distributed AI future where devices, the intelligent edge, on-premises infrastructure, networks and cloud data centers work together as a unified compute fabric. Through real-world demonstrations and ecosystem collaborations, Vinesh will showcase how Qualcomm technologies are helping create the distributed foundation needed to support agentic AI across the compute continuum, unlocking better economics and more scalable AI experiences for businesses and consumers alike.
Real-world AI in action
Visitors to the Qualcomm booth will see firsthand how Qualcomm Dragonfly technologies translate into practical AI deployments:
Running extremely large models on Qualcomm Dragonfly
One featured demonstration highlights the scalability of the Qualcomm Dragonfly AI accelerator architecture by running Kimi-K2.5, a one-trillion-parameter model, on a single Qualcomm Dragonfly AI200 accelerator card. This proof-of-concept showcases memory capacity and system-level design advantages that make it possible to support extremely large models within a compact footprint. Rather than focusing on benchmark comparisons, the demonstration illustrates how Qualcomm Technologies' architecture approaches memory scalability differently to unlock new deployment possibilities for advanced inference workloads.
Bringing security and governance to agentic AI to drive enterprise innovation
As autonomous AI systems become more capable, enterprises face new governance and security challenges. That's why Qualcomm Technologies is collaborating with Confidential Core AI (CCAI) to develop infrastructure designed specifically for agentic AI applications on Qualcomm Dragonfly solutions. Together, the companies are demonstrating a layered approach, combining technologies that help keep AI agents operating within defined policy boundaries and an encryption environment that keeps enterprise know-how and proprietary information under the enterprise's own ownership and control.
The platform is designed to help organizations establish alignment and confinement controls, manage risk and maintain ownership of sensitive enterprise information. By enabling sovereign deployment models, encryption technologies, and infrastructure optimized for AI inference, Qualcomm Technologies and CCAI are helping enterprises adopt agentic AI without sacrificing security, compliance or control over proprietary data.
Token economics: more models per rack
We are also working with Multiverse Computing to show how their CompactifAI model compression and Qualcomm Dragonfly AI infrastructure can work together to improve large-scale inference economics by reducing model memory footprint and bandwidth requirements.
The story is not simply faster inference; it is about making memory the advantage. By reducing how much memory each model requires, operators can fit larger models, support longer contexts and create more workload capacity per accelerator per rack, serving more sessions within the same rack power envelope, and running more models concurrently without expanding infrastructure.
Together, Multiverse and Qualcomm Technologies are reinforcing the conversation toward the metrics that matter most in real deployments: served capacity, rack-scale value proposition, more models per watt, lower cost per token and ultimately better economics for large-scale AI inference deployments.
Visit Qualcomm Technologies at AI Infra Summit
The demos above represent only a portion of what Qualcomm Technologies will be showcasing throughout the event. Whether you're interested in hyperscale AI infrastructure, sovereign AI deployments, networking innovation, custom silicon, memory architecture or agentic AI applications, our experts will be available throughout the conference to discuss the technologies shaping the next era of AI infrastructure.
Stop by the Qualcomm booth 206 to explore the full Qualcomm Dragonfly portfolio, see live demonstrations and engage directly with the engineers and leaders building the future of AI inference. We also encourage developers and innovators to check out the Qualcomm Technologies-sponsored AI Infra Summit Hackathon where you can learn more about our AI software offerings including tools, framework integrations and model optimizations. You will also have an opportunity to build production-grade AI applications on a variety of Qualcomm platforms, including the latest AI PCs with Snapdragon processors and Arduino UNO Q boards.
Join the team building what's next
As Qualcomm Technologies expands its data center business, we are actively looking for engineers, architects, researchers and innovators who want to help solve some of the industry's hardest challenges in AI, compute, memory, networking and system design. If you're passionate about rethinking infrastructure for the agentic AI era and helping build technologies that can power the next generation of intelligent systems, we'd love to meet you.
The infrastructure revolution is underway. At AI Infra Summit 2026, we're excited to show what's possible when efficiency, intelligence and scale come together.
Go Deeper
What is Qualcomm Dragonfly, and how does it fit into the shift toward agentic AI?
As AI moves from generating one-off responses to continuously reasoning and acting, the infrastructure underneath has to change with it. Qualcomm Dragonfly is our answer: a rack-scale portfolio spanning IP, silicon, cards and complete rack solutions across CPUs, AI accelerators, custom silicon and connectivity, designed so hyperscalers, cloud providers and enterprises can coordinate compute, memory and connectivity as one system.
What makes Qualcomm Technologies’ approach to data center AI inference different from the status quo?
The industry has long optimized infrastructure for peak compute throughput, but agentic AI changes the math: what matters now is how much useful output a system can sustain within a fixed power budget. What sets our approach apart is that we bring together leading-edge AI, high-performance low-power computing and unrivalled connectivity as one coordinated stack rather than discrete parts, reorienting the whole system around efficiency and useful output instead of raw FLOPS.
