On the final day of GTC 2026 in San Jose, NVIDIA announced the general availability of IGX Thor, a Blackwell-based edge AI platform built for safety-critical environments where cloud latency is not an option. The platform is now shipping to partners and customers across healthcare, manufacturing, robotics, transportation, and space exploration.
IGX Thor represents NVIDIA's most aggressive push yet into physical AI at the edge — bringing the same generative reasoning capabilities powering data center workloads into operating rooms, factory floors, and autonomous vehicles.
8x the Performance of Its Predecessor
Powered by the NVIDIA Blackwell architecture, IGX Thor delivers up to 8x the AI compute performance of its predecessor, IGX Orin. The platform features both an integrated GPU and an optional discrete GPU, reaching up to 5,581 FP4 teraflops of AI compute with 400 GbE connectivity.
That level of performance allows developers to run multiple generative AI models simultaneously at the edge — enabling machines that can perceive their environment through multimodal sensor fusion, reason about what they observe, and act in real time.
From Surgery to Space
The breadth of early adopters illustrates the platform's versatility:
- Johnson & Johnson is using IGX Thor to power its Polyphonic digital surgery platform, bringing real-time AI inference directly into the operating room.
- Caterpillar is developing an in-cabin conversational AI assistant to enhance worker productivity and safety on construction sites.
- Hitachi Rail is deploying advanced predictive maintenance and autonomous inspection systems across rail networks.
- KION Group is using IGX Thor alongside NVIDIA's Halos Outside-In Safety workflow to enable AI-augmented perception for warehouse robots, using infrastructure-mounted cameras and dynamic virtual safety fences.
- Planet Labs is processing terabytes of satellite data in orbit, transforming raw imagery into actionable intelligence at lower cost.
- CERN researchers are running physics-inspired AI models that process massive data streams at high throughput.
Medical device companies Barco, Cosmo, and XRlabs are also adopting the platform to build off-the-shelf, medical-certified edge AI solutions. KARL STORZ is developing next-generation endoscopy tools powered by IGX Thor for more accurate diagnoses.
Two Production-Ready Systems
NVIDIA is shipping two configurations: the IGX T5000 module for embedded systems requiring functional safety certification, and the IGX T7000 board kit for high-performance workstations. Developer kits are available now from worldwide distribution partners.
The Bigger Picture
IGX Thor arrives at a moment when the AI industry is pivoting from cloud-first model training to real-world deployment. Jensen Huang has repeatedly emphasized that the next frontier for AI is the physical world — robots, autonomous vehicles, medical devices, and industrial systems that must operate with real-time reliability.
With IGX Thor now generally available, NVIDIA is betting that the same architecture powering trillion-dollar data center buildouts can be miniaturized for the factory floor — and that the companies deploying it first will define the next era of industrial automation.
