Overview
- Nvidia’s four‑day GTC conference runs March 16–19 in San Jose, with Jensen Huang’s Monday keynote set to outline advances across accelerated compute, AI factories, open models, agentic systems, and physical AI.
- Nvidia will showcase how Groq’s ultra‑fast inference technology integrates into its CUDA ecosystem following a multibillion‑dollar deal and talent integration, according to company comments reported by Reuters.
- Media reports point to potential reveals such as an open‑source enterprise agent platform dubbed NemoClaw and a new inference‑focused chip, though both remain unconfirmed.
- Analysts expect roadmap detail spanning Vera Rubin ramp timing and a preview path toward Feynman, plus networking updates tied to co‑packaged optics after supply agreements with Coherent and Lumentum.
- Supply and geopolitical risks are front of mind, including tight HBM, wafer and optics availability, possible export controls, and reporting that Nvidia shifted China‑focused production toward next‑generation platforms.