NVIDIA AI Updates: April 3, 2026

1. NVIDIA Open-Sources Physical AI Data Factory Blueprint for Robotics and Autonomous Systems

NVIDIA. NVIDIA released the Open Physical AI Data Factory Blueprint on GitHub under an open-source license, providing a reference architecture that unifies and automates synthetic training data generation, augmentation, and evaluation for physical AI systems. The blueprint integrates NVIDIA Cosmos open world foundation models and leading coding agents to transform limited real-world training data into large, diverse synthetic datasets, directly addressing one of the core bottlenecks in robot and autonomous vehicle development. The release is positioned to accelerate development cycles for teams building robotics, vision AI agents, and self-driving systems by reducing dependence on expensive physical data collection campaigns. Source

2. NVIDIA and Marvell Announce Strategic Partnership for AI Factory and AI-RAN

NVIDIA. NVIDIA and Marvell Technology announced a strategic partnership on April 2, 2026 to integrate Marvell’s networking silicon and custom ASIC capabilities with NVIDIA’s AI factory architecture and AI-RAN (AI Radio Access Network) platform. The partnership links Marvell’s data center interconnect and custom AI accelerator expertise into NVIDIA’s end-to-end AI infrastructure stack, with initial focus on high-bandwidth connectivity between GPU clusters and AI-RAN deployments in telecommunications. The combination aims to help operators build AI-native wireless network infrastructure where NVIDIA GPUs handle both AI inference and radio processing workloads. Source