What happened
NVIDIA has unveiled the Jetson Orin Nano 2, an edge robotics computer designed for drones, robots, and vision systems. The company positions the board as an entry-level option for running generative AI models directly on a device, rather than in a data centre. According to AI News, the hardware delivers 78 trillion operations per second of AI compute, with 8GB of memory and an eight-core Arm CPU.
NVIDIA claims the new board offers twice the inference performance of the existing Jetson Orin Nano Super while fitting the same compact form factor.
How the room's reading it
The launch is being framed by NVIDIA as a bet on smaller models. The company's argument is that models like Gemma and Qwen have reached an accuracy that makes them useful for real-time reasoning on compact hardware. Early adopters seem to agree — Matic Robots is using the board for its home cleaning robots, and Alphabet’s drone delivery subsidiary Wing is evaluating it for its fleet.
The broader hardware ecosystem is also building around the new board. A long list of partners, including ADLINK and Seeed Studio, are developing carrier boards and systems. The consensus among developers is that this lowers the barrier to entry for local AI, enabling more complex perception and interaction without a constant cloud connection.
Sailfish's take
We see this as less about a single board and more about a shift in the software stack. The hardware is a welcome, incremental improvement. The real story is that NVIDIA is pushing a more coherent software narrative — with its Jetson agent skills — for models that can actually run on this class of device.
This isn't just for home cleaning bots. We think the bigger opportunity is in industrial and agricultural applications where low-power, real-time perception is critical. The board makes deployment more accessible, but the hard work for builders remains the same — fine-tuning memory-efficient models for specific physical tasks. The useful question isn't about TOPS. It's whether the agent stack is mature enough to build reliable products on.