NVIDIA Jetson Orin Nano 2 delivers 78 TOPS for edge AI

The introduction of 78 TOPS at the entry level significantly lowers the barrier for deploying complex vision-language models in autonomous hardware. By doubling inference performance within the same compact form factor, NVIDIA enables streaming-adjacent technologies like drone-based delivery and smart camera systems to process high-bandwidth visual data locally rather than relying on cloud compute. This shift addresses latency and privacy concerns critical for real-time spatial awareness in the robotics ecosystem. As partners like AAEON and ADLINK build out the supporting hardware stack, the industry should monitor the H1 2027 release for shifts in how edge devices handle multimodal generative AI workloads.

NVIDIA's Jetson platform continues to attract a broad ecosystem of hardware partners building edge AI systems for robotics and autonomous applications. In August 2025, NVIDIA announced that the Jetson Orin platform had surpassed 2 million cumulative units shipped across industrial, robotics, and automotive deployments, reinforcing the scale of the installed base that the Orin Nano 2 will inherit. Partners such as AAEON, ADLINK, and Advantech have each released carrier boards and system-level products targeting the Jetson Orin family, and Seeed Studio launched its reComputer J4012 edge AI system built on Jetson Orin NX in early 2025, targeting smart factory and logistics use cases. The breadth of this partner network means the Orin Nano 2's 78 TOPS specification will likely reach production hardware across dozens of form factors within months of its H1 2027 availability.

On the business and licensing side, NVIDIA has positioned the Jetson Orin Nano 2 as a bridge to its broader generative AI software stack. The module supports NVIDIA Cosmos world foundation models and Nemotron language models, both of which NVIDIA made available under open licenses at GTC 2025 in March, reducing per-unit software costs for robotics developers. This open-licensing strategy contrasts with competitors like Qualcomm, whose Qualcomm Robotics RB6 platform targets similar edge inference workloads but relies on proprietary Snapdragon software tooling. The pricing and licensing model for Jetson modules has historically been a key differentiator for NVIDIA in winning design slots with mid-tier robotics OEMs, and the Orin Nano 2's entry-level positioning extends that approach to cost-sensitive applications such as delivery drones and warehouse automation.

Independent benchmarks and adjacent deployments highlight the performance trajectory of the Jetson Orin family. In testing published by MLPerf Inference v4.1 results from March 2025, Jetson Orin modules achieved leading edge-class scores in object detection and image classification tasks, outperforming comparable x86 and ARM-based edge accelerators on a per-watt basis. Meanwhile, Wing, the Alphabet-backed drone delivery company, confirmed in 2025 that its autonomous delivery fleet uses NVIDIA Jetson Orin modules for onboard perception and path planning, processing real-time video feeds at the edge without cloud round-trips. These deployments demonstrate the production readiness of the Orin architecture in latency-critical vision workloads, and the Orin Nano 2's doubled inference throughput at lower power consumption positions it to serve similar applications at a lower price point when it ships in 2027.