The AI Grid presents excerpts from Nvidia’s GTC Taipei announcements, including Nemotron 3 Ultra and the Vera CPU for agent workloads. Performance and cost comparisons are Nvidia’s claims.
The video also covers Cosmos 3 for physical AI and RTX Spark, a platform combining CPU and GPU components for local AI computingOn-device inference runs a trained model on the user's device rather than relying entirely on a remote model service.. Jensen Huang appears in the keynote excerpts.
The announcements and product descriptions are supported by Nvidia’s GTC Taipei coverage; the video’s narrator is not identified in the reviewed evidence.
Watch on YouTube




