NVIDIA Unveils AI Factory Reference Design with Industry Partners

NVIDIA has announced a groundbreaking initiative aimed at revolutionizing traditional data centers into advanced AI factories, as revealed at the AI Infrastructure Summit in Silicon Valley. This initiative, led by Ian Buck, NVIDIA’s VP of Accelerated Computing, involves the development of reference designs that will be shared globally with partners and enterprises, providing a blueprint for creating high-performance, energy-efficient AI infrastructure.

Transforming Data Centers

NVIDIA’s ambitious plan is to collaborate with a host of companies across the technology stack, from building design and grid integration to power and cooling solutions. This marks a significant evolution for NVIDIA, expanding its focus beyond chips and systems to encompass a new class of industrial products. The complexity of these AI factories necessitates collaboration, as no single entity can construct them alone.

Jacobs, acting as the design integrator, plays a crucial role in coordinating the physical and digital layers to ensure seamless orchestration. A key feature of the reference design is the creation of a digital twin for the AI factory, integrating IT systems with operational technology for power and cooling both inside and outside the data center.

Collaboration and Innovation

Longtime partners like Schneider Electric, Siemens Energy, and Vertiv have been instrumental in developing resilient, high-efficiency environments tailored for AI-scale workloads. Siemens Energy contributes significantly to on-premises power delivery, meeting the extensive energy demands of these facilities. GE Vernova also plays a part in power generation and electrification.

The initiative extends the digital twin concept to include local power generation, energy storage systems, and AI agents for operations. This comprehensive approach ensures that every watt of energy contributes directly to intelligence generation, optimizing the entire system from power distribution to cooling and system orchestration.

System-Level Optimization

NVIDIA’s strategy involves designing infrastructure and technology stacks in tandem, enabling true system-level optimization. This approach integrates power, cooling, compute, and software into a unified whole. Simulation technology is pivotal in this shift, allowing companies to share simulation-ready assets and model components using AI factory digital twins before physical availability.

These digital twins optimize AI factories pre-construction and aid in management post-deployment. By adopting the OpenUSD framework, the simulation platform can model every aspect of facility operations accurately, fostering smarter, more reliable facilities.

Expanding the Ecosystem

Earlier this year, NVIDIA introduced the Omniverse Blueprint for AI factory digital twins, connecting platforms like Cadence and ETAP. Recent expansions include integrations with Jacobs and Siemens, enabling unified simulation of power, cooling, and networking systems. Once completed next year, this blueprint will allow partners to engage with the system through APIs and digital assets, facilitating real-time collaboration and orchestration across the entire lifecycle.

As a result, AI factories will be designed for composability, resilience, and scale, contrasting with traditional isolated facilities. For further insights, visit the NVIDIA Blog.

Image source: Shutterstock