Supermicro

Built to Accelerate AI

As AI systems scale from megawatts to gigawatts, the center of gravity is shifting from models to infrastructure. Join three leaders operating at the intersection of compute, systems, and national platforms as they examine how power, cooling, architecture, and deployment choices now determine performance, sovereignty, and control of intelligence. Users can understand why the primary bottlenecks in advanced AI systems have moved from algorithms to physical infrastructure, including power density, cooling, and latency; learn what it takes to deploy AI infrastructure at gigawatt scale without compromising reliability, security, or economic viability; explore how system architecture and integration choices shape determinism, isolation, and trust in large-scale AI environments; and gain insight into how sovereign and enterprise AI deployments differ from traditional hyperscale cloud models in governance and control.

Enterprises are moving from AI experimentation to production, where ROI matters as much as performance. This session presented by Supermicro and NVIDIA explores how AI factories, edge AI, and on-prem deployments work together to deliver scalable, cost-effective AI solutions. We will highlight how systems powered by GPUs such as the NVIDIA RTX PRO 6000/4500 Blackwell Server Edition support inference, visualization, and Retrieval-Augmented Generation (RAG) workloads across data centers and edge environments. The session will also examine how RAG pipelines benefit from tightly integrated GPUs, networking, and storage to deliver low-latency, context-aware results—helping enterprises improve data control, reduce operational costs, and maximize ROI from AI investments.

As AI factories scale to support larger models and higher cluster densities, liquid cooling has become essential. This session presented by Supermicro and NVIDIA explores how next-generation AI factories are built using liquid-cooled NVIDIA HGX systems and NVL72 rack-scale platforms, enabling dense, high-performance GPU clusters for large-scale AI training. Attendees will gain insight into Supermicro's deployment of these platforms as validated, rack-scale solutions rather than standalone servers. The session will cover key design and deployment considerations, including direct-to-chip cooling and rack-level liquid integration, and show how liquid-cooled total solutions accelerate time-to-online while improving performance and energy efficiency for modern AI factories.