Building AI inference that scales: Inside the Qualcomm AI200 Rack, Card and AI Infrastructure Management Suite

What you should know:

• AI's integration into data center means service providers balance scale, efficiency and operational complexity to support growing AI workloads.
• Qualcomm Technologies is demonstrating a rack-level AI inference system at MWC 2026, integrating acceleration, memory, interconnect and management software into a cohesive, deployment-ready platform.
• Hardware, connectivity and software come together as a single data center platform, designed to scale with customers as AI workloads evolve.

AI's impact on the data center is no longer theoretical. Model complexity and processing volume keep growing, deployment patterns are shifting, and service providers are being asked to find a delicate balance among scale, efficiency and operational complexity to compete and sustain profitability. At Qualcomm Technologies, our focus has been to approach this moment with intent - applying proven system-level strengths to the evolving requirements of AI inference infrastructure.

Over the past year, we've continued to bring together key building blocks for the data center:

• high-performance, energy-efficient AI acceleration;
• rack-level system design; and
• the software required to deploy and manage these environments at scale.

Designed for sustained operation, reliability and scale, this same system-level approach is foundational to our broader evolution in industrial and infrastructure computing. At MWC 2026, we'll be sharing tangible progress across each of these areas through demonstrations in our booth.

A closer look at the Qualcomm AI200 Rack

One of the centerpieces will be our Qualcomm AI200 rack, integrating accelerator cards, memory architecture, interconnect and management software into a cohesive, ready-to-deploy system. This rack-level approach reflects how customers increasingly evaluate AI infrastructure - not as isolated components, but as complete, serviceable systems designed for sustained operation. The Qualcomm AI200 rack offers a groundbreaking memory capacity of 43 TB, making it ideal for running inference using the latest and largest flagship AI models. The Qualcomm AI200 racks will begin deployment this year, demonstrating how Qualcomm Technologies solves the compute and connectivity bottlenecks, not just at the edge, but now in the core of data centers.

We'll also offer a demonstration of a 350-billion-parameter generative AI model running on a single Qualcomm AI200 card, showcasing the scale that can be achieved today on a single accelerator. The Qualcomm AI200 platform is designed to support models scaling up to 1 trillion parameters,¹ highlighting the importance of system balance - memory capacity, data movement and efficiency working together to deliver real-world generative AI at massive scale.

Connectivity and orchestration at system scale

Equally important is what connects and orchestrates these systems. In December, the acquisition of Alphawave Semi brought an array of core technologies including, but not limited to, high-speed wired connectivity, custom silicon and chiplet technologies into Qualcomm Technologies' data center portfolio. This expertise in high-performance, low-power data movement complements our AI and compute platforms, strengthening our ability to address the growing demands of AI workloads at the system level.

At MWC, this integration comes to life through our Qualcomm AI Infrastructure Management Suite, which HUMAIN is deploying now in data centers. 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.

Qualcomm Technologies' approach to the data center is intentional and grounded in execution - bringing together AI acceleration, connectivity and software into platforms designed for real deployment. MWC is an opportunity to demonstrate progress in working systems. We look forward to providing more information, including an update on our roadmap at our next investor event, where we'll have more to share.