On October 12, at the OCP Global Summit, NVIDIA unveiled a host of pivotal initiatives. It made the design of the Vera Rubin rack and compute tray, along with the NVIDIA MGX architecture, publicly available and launched version 1.2 of the specification. This specification can accommodate rack deployments with a power capacity of up to 250kW, facilitating a more streamlined operational deployment process. Furthermore, NVIDIA introduced four open-source software components tailored for the DSX AI Factory. These components are specifically crafted to aid cloud service providers in swiftly accomplishing hardware deployments for production clusters. NVIDIA also forged alliances with a number of enterprises, such as Cisco, Google, Meta, and Microsoft, to collectively release open standards for AI infrastructure. These standards encompass unified specifications for the power supply, cooling systems, and physical layout of AI factories. They also outline grid resilience requirements for large-scale AI facilities, establish pluggable optical module standards for 400Gb/s network architectures, and offer solutions to boost the transmission efficiency of Ethernet AI networks. Meta declared its intention to utilize NVIDIA's relevant switch products to construct a Vera Rubin AI factory capable of supporting over 100,000 GPUs.