Why Deploying Physical AI at Scale Demands Safety at Every Layer
NVIDIA says scaling physical AI requires layered safety and promotes its Halos stack for vehicles and robots.
NVIDIA argues that autonomous vehicles and robots need safety across hardware, software, AI behavior, and operating environments as physical AI scales. ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles by 2035, while Omdia estimates about 60 million industrial robots deployed between 2026 and 2035. The company presents NVIDIA Halos as a full-stack safety system spanning DRIVE AGX Thor, Hyperion, Halos OS on ASIL-D DriveOS, Alpamayo vision-language-action models, IGX Thor, Holoscan, Isaac Lab, and Omniverse. It cites emerging ISO/IEC TS 22440 guidance and an AI Systems Inspection Lab intended to support third-party certification.
- ABI projects 49 million level 3-5 AVs by 2035.
- Omdia estimates 60 million industrial robots from 2026-2035.
- Halos covers compute, OS, models, sensing, and simulation.
- ISO/IEC TS 22440 is cited for AI-specific assurance.
- Vendors
- NVIDIA
- Products
- NVIDIA HalosNVIDIA DRIVE AGX ThorNVIDIA HyperionDriveOSNVIDIA IGX ThorNVIDIA HoloscanNVIDIA Isaac LabNVIDIA Omniverse
- Organizations
- ABI ResearchOmdia
- AI models
- Alpamayo
From the original article
5 min read at blogs.nvidia.com
Physical AI is moving rapidly from research to large-scale deployment. By 2035, ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles (AVs), while Omdia estimates that roughly 60 million industrial robots will be deployed between 2026 and 2035. As these machines enter roads, factories, warehouses and other environments shared with people, […]
By Riccardo Mariani · blogs.nvidia.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.