The concept of the AI factory has emerged as the defining infrastructure of the modern intelligence era, representing a fundamental shift from isolated hardware components to fully integrated industrial systems. Unlike traditional setups, these factories rely on extreme codesign, where hardware, networking, memory, storage, and software are architected together. This holistic approach ensures continuous optimization at every layer, aiming to maximize utilization, lower the cost per token, and significantly boost output. The primary goal is to balance the need for real-time responsiveness required by always-on interactive AI workloads with the massive throughput necessary for high-volume production. As AI workflows become longer and more interactive, inference has evolved into a real-time orchestration challenge. The factory must dynamically route requests, manage memory, coordinate services, and balance latency against throughput while maintaining high utilization across the entire stack. In this context, the software layer is critical, as the efficiency of the orchestration directly determines the volume of intelligence produced and the value generated. This challenge spans the full machine, requiring a live, adaptive management system. What began as a focus on graphics processing units has expanded into full-stack AI factories. These systems now comprise accelerated compute, high-speed interconnects, liquid-cooled environments, advanced inference software, autonomous agents, and reference architectures. NVIDIA is leading the definition and construction of this ecosystem in close collaboration with global system partners including Cisco, Dell, HPE, Lenovo, and Supermicro. Together, they are delivering comprehensive AI infrastructure to enterprise data centers. Furthermore, a curated ecosystem of AI software partners enables the creation of tailored solutions for specific enterprise use cases, offering flexibility across both proprietary and open-source model options. These AI factories are versatile, supporting a wide array of applications ranging from agentic AI and physical AI to robotics. Every organization across all industries, from financial services and life sciences to manufacturing and the public sector, will need to either build or rent access to such facilities. As a proof of concept, NVIDIA itself operates an enterprise AI factory to accelerate its internal development. This facility employs hundreds of autonomous AI agents to assist engineering, software, and operations teams, demonstrating how AI factories can transform corporate productivity by weaving AI capabilities directly into daily workflows. Deployment can range from small-scale implementations supporting a single business unit to massive gigawatt-scale constructions designed for high-performance inference and training. To achieve this scale, NVIDIA utilizes DSX reference designs to unify design, simulation, operations, and ecosystem technologies. Building these facilities requires more than just optimized compute; it demands a shared digital environment where facility design, power, cooling, and operations are modeled together before construction and continuously improved after deployment. The NVIDIA Omniverse DSX Blueprint supports this by creating digital twins that connect facilities, hardware, and software, allowing partners to validate designs and optimize operations throughout the lifecycle of the AI factory. Ultimately, a full-stack approach allows organizations to extract greater intelligence from every system, transforming AI infrastructure into an autonomous engine of reasoning and insight. Just as the last industrial revolution converted energy into work, the current era converts energy into intelligence. AI factories serve as the essential infrastructure for this transition, built to power the next wave of global economic growth.