Nvidia GTC: new robotics, physics, chips
Nvidia kicked off GTC 2026 with promises of next‑gen GPUs, a real‑time physics engine called Newton, and an open humanoid robot foundation model dubbed Isaac Groot N1 — all billed to accelerate robotics R&D and agentic AI workflows live coverage and Reuters. The keynote signaled a move to more specialized AI chips for real‑world systems — think simulation + control, not just big‑model training — which shifts what demo projects and infra skills will stand out this year.
Newton was contributed(linuxfoundation.org) to the Linux Foundation as an open-source, GPU‑accelerated physics engine co‑developed by NVIDIA, Google DeepMind and Disney Research. The engine is built on NVIDIA Warp and uses OpenUSD for scene interchange(developer.nvidia.com). The Newton GitHub repository shows roughly 2.6k stars and about 1,538 commits, with active merges and recent PR activity over the past weeks(github.com). NVIDIA’s GR00T lineage advanced to GR00T N1.6, published on Hugging Face and NVIDIA’s GitHub, with the N1.6 pretraining run reported at 300,000 steps and a global batch size of 16,384 during pretraining(huggingface.co). Company documentation and the March 18, 2025 press release describe GR00T’s dual‑system VLA design—an explicit “System 2” vision‑language planner paired with a high‑frequency action Diffusion Transformer—and NVIDIA has signaled N1.6 integrates its Cosmos Reason component for improved robot reasoning(nvidianews.nvidia.com). NVIDIA’s Rubin platform papertrail (press release, Jan 5, 2026) claims the Vera Rubin codesign (Vera CPU, Rubin GPU, NVLink 6, ConnectX‑9, BlueField‑4 DPU) can cut inference token cost up to 10× and reduce required GPUs for MoE training by 4× versus Blackwell(investor.nvidia.com); CNBC reported Vera Rubin is slated to roll out during 2026 for broader deployment planning(cnbc.com). Multiple outlets reported NVIDIA is simultaneously positioning NemoClaw (aka NemoClaw/NemoClaw‑branded agent stack) as an open‑source enterprise AI agent platform intended for hardware‑agnostic deployment across cloud and edge environments(cnbc.com).
