Nvidia unveiled Spectrum-6, a 102.4-terabit-per-second Ethernet switch system that doubles the capacity of its predecessor and is a core component of the Vera Rubin AI supercomputer framework designed for large-scale, agentic AI workloads. Early adopters and customers include CoreWeave, Microsoft, Nebius, SpaceXAI, and Tesla, with several partners planning to deploy Vera Rubin-based infrastructure that aggregates more compute into a single pool for faster model training and large-scale inference. The rollout prompted about a 2% rise in Nvidia’s stock as investors gauge the potential impact on AI data-center efficiency and cost. AMD also announced Microsoft as a customer for its rival Helios Rackscale Solution, joining Meta, OpenAI, and Oracle in sign-ups, highlighting competition in the cloud-scale AI infrastructure space. The Spectrum-6 effort is positioned as a stepping stone toward the broader Spectrum-X platform, emphasizing higher bandwidth, improved traffic management, and fault tolerance for AI factories.