From GPU companies to system companies, NVIDIA's transformation narrative has been further refreshed, and the infrastructure war has entered the second half.

View Original
BlockBeatNews
Jensen Huang explains in detail why Vera Rubin was born: The Agent era requires a brand-new heterogeneous computing foundation
Huang Renxun at GTC Taipei 2026 described Vera Rubin as an end-to-end system—not just a chip or GPU. The full stack includes Rubin GPU, NVLink, Vera CPU, ConnectX-9, BlueField-4 DPU, and DOCA. Agents handle GPU inference, CPU scheduling, and cross-CPU/GPU tool calls, with end-to-end encryption throughout enabled by the BlueField DPU. The memory system is innovating storage through KV cache and retrieval. CUDA X will have Skills, and Agents will be able to learn how to call libraries. NVIDIA is transforming from a GPU company into a systems company, building AI factory-level infrastructure.
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned