日本語 
NVIDIA has shared a reference design for generation AI data centers. This design gives guidance on how to build systems that use NVIDIA GPUs. It’s meant to help data-center operators and tech companies create facilities that can handle the rising power and computing needs of AI tasks.
The design covers parts of the infrastructure needed for big AI systems. That includes power delivery, cooling, networking, computing and how everything fits together inside racks. By offering an approach NVIDIA wants to give its partners a solid foundation for building and launching AI infrastructure. This is especially important as GPU clusters get denser and more demanding.
This move comes at a time when AI and AI agents are pushing up demand for powerful computing. Old-style data centers often run into issues with power density, heat and fast connections when handling groups of GPUs. These problems are making companies rethink how they design data centers of just adding more servers the old way.
Also Read: Mavenir, Neysa Advance Sovereign AI Infrastructure
For businesses and infrastructure providers having a reference design makes planning AI-ready facilities easier. It helps teams work better together across hardware and data-center parts. This also shows a trend in the industry. Moving toward AI-specific infrastructure. In these setups computing, networking, power and cooling are all planned as one connected system.
As AI tasks grow bigger the way data centers are built is becoming a part of the AI supply chain. Things like efficiency, scalability and stable power delivery are now priorities, in infrastructure planning.
English