As frontier LLMs grow ever larger, the challenge is no longer just raw compute—it’s deploying them efficiently, reliably, and at scale. In this session, Rebellions presents its full-stack approach to AI infrastructure, from advanced NPUs and chiplet interconnects to an optimized software stack and production-ready deployment frameworks. Attendees will see a demo of MoE inference and explore real customer use cases spanning security, vision, and enterprise AI applications. Discover how Rebellions’ approach bridges innovation and practicality—delivering enterprise-grade efficiency, scalability, and lower TCO for real-world data centers.

As AI-native cloud platforms scale to meet global demand, the data center IT infrastructure powering generative AI has become a prime target for attackers. In this session, Dr. Yuriy Bulygin—CEO of Eclypsium and former Chief Threat Researcher at Intel—shares a case study from one of the world’s fastest-growing AI cloud providers that supports OpenAI, Microsoft, and NVIDIA workloads and how they are offering secure AI infrastructure that their customers demand.

This provider faced the challenge of protecting thousands of specialized servers, GPUs, and system components—without slowing growth. Learn how they implemented a turn-key approach to data center security, leveraging firmware integrity verification, automated vulnerability management, and continuous supply chain monitoring.

Attendees will gain insight into:

• Why infrastructure threats in AI data centers are rising—and often invisible
  
  

• What does a secure AI cloud look like and how to minimize risk
  
  

• Lessons for building resilient, secure AI data centers without adding operational drag
  
  

If you're responsible for AI infrastructure, securing the data center isn’t just an IT concern—it’s foundational to model integrity and platform trust.

California faces over 7,000 wildfires each year, with enormous costs to lives, communities, and ecosystems. Responding faster requires distributed sensing and intelligence that can act in the field where traditional satellites and watchtowers fall short. http://Wywa.ai First Responder is an open-science initiative led with researchers from MIT and CMU, together with industry leaders and policy experts, to design and deploy a scalable wildfire early-warning network. The system combines ultra-low-cost LoRa-enabled chemical sensors with edge AI and vision-language models. These distributed “artificial noses” continuously monitor air for smoke and combustion signatures. When risk thresholds are detected, the sensors activate nearby edge vision systems that confirm wildfire presence and generate real-time alerts for first responders and civic authorities. We will present results from early deployments, highlight the LoRa network architecture and AI model training that make such systems deployable at scale, and discuss how open collaboration across academia, industry, and government can accelerate resilience. The session will include a live demonstration of how edge intelligence can empower communities to act in the earliest, most critical moments of wildfire response.

What does it take to run one of the world's largest AI supercomputers? As artificial intelligence workloads grow exponentially, operating a hyperscale AI cloud fleet demands new strategies for resilience, efficiency, and operational excellence. This session explores Microsoft’s approach to scaling infrastructure for 100X growth, focusing on the intersection of system innovation and advanced fleet management.