The year 2025 saw significant developments in the world of GPU architecture, with the Blackwell and Rubin architectures leading the way into the era of large AI workloads.
Nvidia and Cisco deepened their collaboration to bring AI infrastructure to enterprise settings, introducing the Cisco Nexus HyperFabric with Nvidia AI to streamline the deployment of AI clusters for standard IT teams.
Nvidia’s Spectrum-X Ethernet platform gained traction in 2025, highlighting Ethernet’s suitability for the mass enterprise market and bringing advanced capabilities like remote direct memory access (RDMA) to standard Ethernet for AI workloads in data center networking. Additionally, the US government’s approval of H200 chips for China marked a pivotal moment in US-China tech policy and supply chain dynamics.
