Nvidia debuts Jetson AGX Thor edge AI platform for robotics

Nvidia has unveiled its most powerful Jetson edge platform to date, signaling a major push into robot intelligence with the Jetson AGX Thor. The centerpiece is the Jetson T5000 system-on-module, paired with the new AGX Thor developer kit, delivering a claimed 2,070 teraflops of 4-bit floating point performance while operating within a 130-watt envelope. The kit positions developers to build and test sophisticated AI and robotics applications at the edge, including humanoid robotics, autonomous machines, and sensor-heavy systems.

At the heart of the Thor platform is Nvidia’s Blackwell GPU architecture, complemented by a 14-core Arm Neoverse V3AE CPU and 128 gigabytes of LPDDR5X memory. The configuration enables high-throughput AI workloads and real-time processing across multiple sensors and inputs. The system supports four 25-gigabit Ethernet ports and a high-bandwidth NVMe storage option via PCIe, with a reference carrier board that includes a 1-terabyte NVMe SSD. For video workloads, the platform can encode and decode multiple 4K and 8K streams, a capability Nvidia says is essential for visual AI applications and sophisticated robot perception.