NVIDIA (and its many partners) used its GTC conference to announce a whopping 38 products, as well as to highlight designs, capabilities and partner integrations.
Altogether, these aim to deliver a new infrastructure stack for agentic and physical artificial intelligence, spanning chips, racks, software and industry alliances - all accordig to NVIDIA, and its burst of announcements that ran alongside GTC.
The 'big tech' company introduced the Vera Rubin platform — a POD-scale architecture that combines seven new chips into NVL72 GPU racks, Vera CPU racks, BlueField-4 STX storage racks and Spectrum-6 Ethernet — and published the Vera Rubin DSX AI Factory reference design and Omniverse DSX digital twin blueprint to guide large-scale AI factory buildouts.
NVIDIA also moved core inference and agent orchestration to production: Dynamo 1.0 is released as an open source inference operating system that can boost NVIDIA Blackwell GPU inference by up to 7x, and the Agent Toolkit, Nemotron open models and NemoClaw for OpenClaw aim to accelerate local and always-on agents.
The company announced the Nemotron Coalition to co-develop frontier open models and expanded Nemotron 3 families, while NemoClaw bundles OpenShell runtime, Nemotron models and privacy routing for personal AI assistants.
On transportation and networking, NVIDIA is extending DRIVE Hyperion adoption with partners including Hyundai Motor Group, Kia, BYD, Geely, Isuzu and Nissan, and announced that Uber will launch a full-stack NVIDIA-powered robotaxi service across 28 markets by 2028.
Telecom operators and cloud providers are building geographically distributed AI grids and AI-RAN edge infrastructure with AT&T, Comcast and T‑Mobile to run inference closer to users.
Related reading
- Nordic Semiconductor expands 'ultra low power' AI chip lineup
- Apple buys Polish video editing plug-in maker MotionVFX in push to challenge Adobe
- NVIDIA survey finds healthcare AI moving from pilots to production with measurable returns
Industry applications span robotics, healthcare, gaming and space: NVIDIA unveiled new Isaac and Cosmos models for physical AI and a Physical AI Data Factory Blueprint used by FieldAI, Uber and Teradyne Robotics, while Roche is deploying more than 3,500 NVIDIA Blackwell GPUs for drug discovery.
The company also integrated CloudXR with Apple Vision Pro and previewed DLSS 5 for photoreal game rendering. “Physical AI has arrived — every industrial company will become a robotics company,” said Jensen Huang, founder and CEO of NVIDIA.
The recap
- Vera Rubin platform introduces seven chips and POD-scale stacks
- Roche deploys more than 3,500 NVIDIA Blackwell GPUs globally
- Uber to roll out robotaxis across 28 markets by 2028