The Lambda Hyperplane-16 is a high-performance data center server designed for training large neural networks at scale. It features 16x NVIDIA Tesla V100 SXM3 GPUs, NVSwitch architecture, and InfiniBand cards, allowing for fast GPU-to-GPU communication within the server and across multiple servers during distributed training. The Hyperplane-16 offers several advantages over existing solutions, including more performance and lower cost compared to other 16x GPU servers, significantly shorter training times than 8x GPU servers for small and large models, pre-installed AI frameworks and drivers, and out-of-the-box distributed training capabilities. The server's technical specifications include a 10U rackmountable design with up to 1.5 TB ECC RAM, 16x Tesla V100 SXM3 GPUs, NVSwitch architecture, and InfiniBand cards. The Lambda Hyperplane-16 is designed for research and engineering teams at leading companies and universities, providing enterprise-class support, pre-installed software, and instant quotes for a new machine.