At the GTC 2025 event, NVIDIA made a significant announcement, unveiling its Omniverse Blueprint for AI factory design and operations. The new feature aims to revolutionise the way AI factories — large-scale data centres dedicated to AI training and inference — are designed, tested, and optimised using cutting-edge simulation technology.Shaping the future of AI factories
With AI now a mainstream technology, the demand for AI factories is growing rapidly. These specialised facilities are purpose-built for training artificial intelligence models and processing vast amounts of data. Many of these factories are expected to be built at a gigawatt scale, a monumental engineering challenge.
Constructing a single gigawatt AI factory involves coordinating thousands of workers across various suppliers, architects, contractors, and engineers, all working to assemble billions of components and miles of fibre cable.ALSO READ: NVIDIA GTC 2025: Jensen Huang unveils breakthroughs in AI and computingTo streamline the design and construction of these AI factories, NVIDIA’s new Omniverse Blueprint will allow engineers to plan and simulate every aspect of the process before construction even begins.Simulation-driven design approachThe Omniverse Blueprint integrates NVIDIA’s own advanced computing systems with a suite of powerful simulation tools from partners such as Cadence, ETAP, Schneider Electric, and Vertiv. This allows engineers to aggregate 3D data from multiple sources — including building design, power systems, cooling units, and more — to test, simulate, and optimise every component of an AI factory.NVIDIA founder and CEO Jensen Huang highlighted the benefits of this approach during his keynote at GTC 2025, showcasing how the blueprint can be used to simulate a 1-gigawatt AI factory. With access to leading simulation tools like the Cadence Reality Digital Twin Platform, engineers can evaluate the performance of power, cooling, and networking systems long before construction begins, ensuring a more efficient and effective design.Addressing complex engineering challengesThe Omniverse Blueprint allows engineers to address several challenges typically faced in AI factory design:
Component integration and space optimisation: The blueprint helps to optimise the integration of complex systems like NVIDIA’s DGX SuperPODs, GB300 NVL72 systems, and over 5 billion components, ensuring that space is used effectively.
Cooling system performance: Using the Cadence Reality Digital Twin Platform, the blueprint allows for simulations of hybrid air- and liquid-cooling systems, ensuring optimal cooling efficiency for the large-scale operations of an AI factory.
Power distribution: By integrating ETAP’s simulation capabilities, the blueprint can model scalable, redundant electrical systems, ensuring the efficiency and reliability of power supply in AI factories.
Networking infrastructure: The blueprint optimises networking setups using NVIDIA Spectrum-X and the NVIDIA Air platform, simulating high-bandwidth infrastructure to ensure fast and reliable data flow across the entire facility.
Breaking down engineering silosOne of the biggest challenges in AI factory construction is the siloed approach to different disciplines, such as power, cooling, and networking. These often separate teams can lead to inefficiencies and missed opportunities for optimisation. The Omniverse Blueprint aims to break down these silos by enabling real-time collaboration across disciplines.
Engineers can now work together in a shared virtual environment, testing how changes in one area, like cooling, impact other areas, such as power usage and networking. This collaborative approach ensures that the final design is both efficient and robust, reducing the risk of failure points that could lead to costly downtime.Real-time simulations for rapid decision-makingIn Huang’s demonstration of the Omniverse Blueprint, engineers were able to make real-time adjustments to the factory’s design and immediately see the impact. For example, a small change in the cooling layout resulted in significant improvements in efficiency, which could have been easily overlooked using traditional methods.The key advantage here is speed: what once took hours or even days to simulate can now be tested and refined in seconds. This rapid feedback loop allows engineers to quickly optimise every aspect of the AI factory, from the cooling system to the power grid.Future-proofing AI factoriesAs AI technology continues to evolve, so too will the demands placed on AI factories. The Omniverse Blueprint takes this into account, offering tools to predict how changing AI workloads will impact power, cooling, and networking requirements at scale. It also allows for the simulation of failure scenarios — such as grid failures or cooling leaks — ensuring that the AI factory is resilient to potential risks.Moreover, the blueprint enables users to plan for future expansions and upgrades. It models cost and downtime scenarios, helping operators prepare for changes in the facility’s infrastructure years ahead of time, ensuring that AI factories remain future-proof as AI workloads evolve.Minimising downtime and riskThe economic stakes of AI factory downtime are enormous. For a 1-gigawatt AI factory, even a single day of downtime can result in losses exceeding $100 million. By using the Omniverse Blueprint to solve engineering challenges before they arise, operators can dramatically reduce both the risk of costly downtime and the time required for deployment.The road to agentic AI for AI factory operationsLooking to the future, NVIDIA is working on the next evolution of the Omniverse Blueprint, integrating AI-powered operational systems to enhance the resilience and performance of AI factories. Partners like Vertech and Phaidra are collaborating with NVIDIA to bring AI-enabled solutions to the blueprint.Vertech is working with NVIDIA’s engineering team to develop an advanced AI control system for factories, combining IT and operational technology to improve operational visibility and resilience. Meanwhile, Phaidra is integrating reinforcement-learning AI agents into the Omniverse Blueprint. These agents optimise energy efficiency and thermal stability in real-time, continuously adapting to changing conditions in the factory environment.AI is driving the data centre boomThe rapid rise of AI is transforming the global data centre landscape, with $1 trillion expected to be spent on AI-driven data centre upgrades. As this transformation takes shape, the use of digital twin technology — as seen in the Omniverse Blueprint — is no longer just an option, but an essential tool for the efficient operation of AI factories.The NVIDIA Omniverse Blueprint for AI factory design and operations is set to play a central role in leading this transformation, allowing AI factory operators to stay ahead of evolving workloads, minimise downtime, and maximise efficiency.ALSO READ: GTC 2025 | NVIDIA launches Halos, a safety system for autonomous vehicles
