In a bid to bring real-time “physical AI” directly into industrial floors and medical operating rooms, NVIDIA has unveiled the IGX Thor platform, built on the Blackwell architecture and optimized for industrial-grade, edge-based AI applications. The IGX Thor is positioned as a powerful module that merges high-speed sensor processing, enterprise-grade reliability and functional safety to support robotics, manufacturing automation, medical devices and other physical-AI systems. Compared to its predecessor, IGX Orin, it offers up to 8× the AI compute for integrated GPUs, 2.5× on discrete GPUs, and 2× better connectivity, enabling large language and vision–language models to run at the edge with 5,581 FP4 teraflops of AI compute in a form factor built for long-term industrial deployment. The platform supports a 10-year lifecycle and integrates the NVIDIA AI Enterprise software stack — including NIM microservices, Isaac for robotics, Metropolis for visual AI and Holoscan for sensor processing — along with NVIDIA’s “Halos” functional-safety elements, making it suited for human-robot collaboration and mission-critical environments. Early adopters span industrial, robotic and medical leaders such as Hitachi Rail, Maven Robotics, Diligent Robotics, EndoQuest Robotics and CMR Surgical, and the company partners with a broad ecosystem of hardware vendors to deliver edge servers, carrier boards, cameras and sensors.
Also Read: NVIDIA Advances Open Frameworks to Drive Next-Gen Robotics
The arrival of IGX Thor signals a significant step in accelerating the migration of AI from cloud/data-centers into real-world physical systems, enabling autonomous inspection in rail networks, advanced robotic assistance in logistics and manufacturing, and surgical-robotic capabilities that “process high-fidelity data in real time, allowing us to simplify complex procedures and enable safer, more intelligent minimal-access surgery,” according to Chris Fryer, CTO of CMR Surgical. In the words of Giuseppe Marino, Group CEO of Hitachi Rail, “By adopting NVIDIA IGX Thor, we are bringing the world’s most powerful industrial-grade, real-time AI performance directly to the edge, enabling operators to better optimize their railways and infrastructure.” With availability slated for December via the IGX T5000 module and IGX T7000 board kit, IGX Thor is poised to redefine how AI interfaces with the physical world, driving new levels of automation, safety and intelligence across industrial and healthcare environments.
