In the 2025 GTC conference, NVIDIA introduced an innovative design called "Omniverse DSX Blueprint," specifically tailored for terawatt-scale (1 billion watt) AI data centers. This design is also referred to by NVIDIA as an "AI factory," marking another major advancement in artificial intelligence infrastructure.
The "Omniverse DSX Blueprint" is designed to meet the growing demand for AI computing, from 100 million watts to 1 billion watts. It can efficiently train and run large AI models. The design is based on NVIDIA's Omniverse framework, combining digital twin technology with real-world engineering data to create a unified operational environment. Through this platform, partners can improve efficiency in all stages of planning, building, and optimizing AI data centers, covering aspects such as power supply, cooling, computing, and networking.
To accommodate data centers of different scales, NVIDIA offers two main configuration options: DSX Boost and DSX Flex. DSX Boost is an internal configuration that can reduce power consumption by 30% or increase GPU density without additional physical expansion, thereby significantly enhancing the processing capabilities of AI models. DSX Flex is an external configuration that can connect to local power grids and renewable energy sources, utilizing idle 100-gigawatt grid capacity by dynamically balancing energy supply and demand.
Currently, the Omniverse DSX reference design has been validated at NVIDIA's AI factory research center in Virginia and has provided technical support for multiple real-world projects, including the 2-billion-watt Switch data center in Georgia and the 1.2-billion-watt Stargate project in Texas. NVIDIA's series of innovative initiatives will lay a more solid foundation for future AI computing.
Key Points:
🌐 NVIDIA launched "Omniverse DSX Blueprint," specifically designed for terawatt-scale AI data centers.
⚡ Offers two configuration options: DSX Boost reduces power consumption, while DSX Flex connects to renewable energy.
🏗️ The design has been validated in multiple real-world projects, helping advance AI technology.
