NVIDIA introduced IGX Thor, an industrial-grade platform built to deliver real-time physical Artificial Intelligence at the edge for factories, warehouses and operating rooms. The small-form module is designed to process high-speed sensor data with enterprise reliability and functional safety, providing up to eight times the Artificial Intelligence compute of its predecessor, IGX Orin. Early adopters include Diligent Robotics, EndoQuest Robotics, Hitachi Rail, Joby Aviation, Maven and SETI Institute, while CMR Surgical is evaluating the platform for advanced medical capabilities.
IGX Thor integrates two types of NVIDIA Blackwell GPUs, combining an integrated GPU with a discrete GPU to deliver 5,581 FP4 teraflops of Artificial Intelligence compute and 400 GbE connectivity. Compared with IGX Orin, it offers up to eight times higher Artificial Intelligence compute on integrated GPUs, 2.5 times on discrete GPUs and double the connectivity to run large language and vision language models at the edge. The platform includes a 10-year lifecycle and long-term support for the NVIDIA Artificial Intelligence Enterprise software stack. It supports NVIDIA NIM microservices to accelerate development from cloud to edge, NVIDIA Isaac for robotics, NVIDIA Metropolis for visual Artificial Intelligence and NVIDIA Holoscan for sensor processing, and it integrates elements from the NVIDIA Halos safety system to embed functional safety into robotics, medical and industrial systems.
Adoption is underway across transportation and industrial robotics. Hitachi Rail is deploying IGX Thor for predictive maintenance and autonomous inspection on rail networks to improve reliability and efficiency. “Artificial Intelligence and data are transforming railways,” said Giuseppe Marino, group CEO of Hitachi Rail, highlighting the benefits of bringing industrial-grade, real-time Artificial Intelligence to the edge. Maven Robotics will place IGX Thor at the core of its next-generation industrial robots to pair safety-rated compute with performance for advanced embodied-Artificial Intelligence models, according to CEO Hamza Derbas.
In healthcare, Diligent Robotics and EndoQuest Robotics are adopting IGX Thor, and CMR Surgical is evaluating it to power Artificial Intelligence assistants and real-time surgical guidance that process high-fidelity data during procedures. “Precision and patient safety are at the heart of every procedure,” said Chris Fryer, CTO of CMR Surgical, citing the potential to simplify complex operations and enhance outcomes. NVIDIA’s partner ecosystem of equipment makers, including Advantech, ADLINK, ASRock Rack, Barco, Curtiss-Wright, Dedicated Computing, EIZO Rugged Solutions, Inventec, NexCOBOT, Onyx, WOLF Advanced Technology and YUAN, is delivering IGX Thor-powered edge servers, carrier boards, design services, cameras and sensors. The platform includes two production-ready systems, the IGX T5000 module and IGX T7000 board kit, with production systems and developer kits expected in December.
