US20260223314
2026-07-30
Electricity
H05K7/1492
High-density switching embodiments are designed to optimize network efficiency, reduce latency, and enhance scalability in data centers, particularly for machine learning (ML) and artificial intelligence (AI) applications. These configurations include rail-optimized topologies, interconnect cable systems, and rack designs supporting liquid cooling and high-density switching. Cost-effective copper cabling is utilized in some embodiments for efficient data handling and communication in high-computation environments, leading to improved performance and reduced infrastructure costs.
The disclosure pertains to information handling systems, focusing on data centers and computing centers. As the demand for processing and storing information escalates, there is a need for systems that vary in how data is handled, processed, stored, and communicated. This variety allows systems to be general or tailored for specific uses, such as enterprise data storage or global communications. The increasing complexity of ML/AI models has intensified the demand for robust data centers with extensive processing units and infrastructure.
The rapid development of ML/AI applications has led to a surge in the need for computing and processing resources. Graphics processing units (GPUs) are particularly effective for ML/AI tasks, both in training and inferencing. As these models become more complex, the supporting infrastructure, including network switches and cabling, has also grown more intricate. The physical arrangement, cabling type, and network topology are crucial for minimizing processing delays in these high-demand environments.
The embodiments can be implemented in various formats, such as processes, devices, systems, or methods on tangible computer-readable media. Components described may be separate functional units or integrated into single systems, implemented in software, hardware, or a combination. Connections between components are not limited to direct links; data can be modified or reformatted by intermediary components. The terms used are illustrative and should not restrict the scope of the disclosure.
While the embodiments are described in the context of ML/AI applications or NVL72 systems, the disclosure is adaptable to other equipment and contexts. Features from one embodiment can be combined with others, and the experiments and results provided are illustrative, not limiting. The patent document emphasizes the need for adaptable, efficient data center structures to support high-density, high-computation environments, catering to the growing demands of ML/AI applications.