NVIDIA Jetson Orin Nano 2 Targets Entry-Level Robotics in 2027

NVIDIA has announced its next entry-level Jetson robotics computer

NVIDIA announced Jetson Orin Nano 2 on August 25, 2026 as a new compact edge-AI and robotics computer for developers building robots, drones and vision systems. The module and developer kit are expected to become available in the first half of 2027, so this is an announced upcoming product rather than hardware that developers can buy today.

78 TOPS in the same compact Jetson form factor

NVIDIA says Jetson Orin Nano 2 provides 78 trillion operations per second of AI compute, 8GB of memory and an 8-core Arm CPU. The company says improved Tensor Cores and higher memory bandwidth deliver 2x the inference performance of Jetson Orin Nano Super while retaining the same compact form factor.

In 15-watt mode, NVIDIA also says the new module uses 40% less power to deliver the same performance as its predecessor. Those figures are vendor-reported product comparisons and should not be treated as universal real-world results across every model or robotics workload.

Smaller multimodal models are moving onto edge devices

NVIDIA is positioning the module for local inference with compact large language and vision-language models. The company lists optimized open models including NVIDIA Cosmos and Nemotron, Gemma 4 and Qwen 3 among the software developers can target through the Jetson ecosystem.

That direction matters for robotics because local inference can reduce dependence on continuous cloud connectivity and lower round-trip latency for perception, language understanding and decision support. It can also help developers keep some sensor data on the device, although application-level privacy still depends on how the full system is designed.

Early robotics users are exploring home robots, drones and vision AI

NVIDIA says more than 3 million developers use its robotics stack. It names Cognex, Doosan Bobcat and Matic among early adopters or evaluators of Jetson Orin Nano 2. Alphabet subsidiary Wing plans to evaluate the new platform for delivery-drone perception and reasoning, while Matic is adopting it for home-cleaning robots that combine conversational interaction, gesture detection, mapping and semantic understanding.

A broader hardware ecosystem is also preparing carrier boards, systems and reference designs around the module.

Why the product is notable

The most important part of the announcement is not simply the TOPS number. Entry-level robotics platforms increasingly need to run multimodal and generative models locally while staying inside strict power, thermal and size limits. Jetson Orin Nano 2 is NVIDIA's attempt to move that capability further down the cost and power envelope.

Developers evaluating it should compare model-specific latency, memory pressure, sustained thermal behavior, power draw and software compatibility once production hardware becomes available. Until then, NVIDIA's performance figures should be understood as launch specifications and vendor benchmarks for an upcoming 2027 product.