Nvidia has unveiled a new $249 version of its Jetson computer designed for running artificial applications locally. The compact Orin Nano is touted to offer double the speed and efficiency of its predecessor at half the cost, with the ability to handle approximately 70% more computational tasks, as per Nvidia.
The Orin Nano is a perfect fit for hobbyists keen on training their own AI applications or for developers of robots and industrial tools seeking to run advanced applications without relying on cloud connectivity.
In a quirky YouTube video introducing the product, Nvidia CEO Jensen Huang showcases the palm-sized computer by pulling it out of his oven in his kitchen. Huang boasts that the computer can process nearly “seventy trillion” operations per second while consuming just 25 watts of power.
Huang reminisces, “A long time ago, we created a brand new type of processor, it was a robotics processor. Nobody understood what we were building at the time, and we imagined that someday these deep learning models would evolve and we would have robots for everything.” Originally intended to power robots, Jetson computers can now also handle large language models like Meta’s Llama.
Sam Altman’s startup World (formerly Worldcoin), which focuses on authenticating human identity through iris scans, utilizes a Jetpack module inside its Orb scanning device. In a blog post last October, World highlighted the enhanced AI performance of the latest Orb version equipped with an advanced NVIDIA Jetson module.
Essentially, the Orin Nano serves as a portable brain that can be connected to other hardware to power its AI capabilities. While cloud hyperscalers like Amazon and Google charge for server access and AI models, the Orin Nano offers a cost-effective solution for running lightweight AI applications without relying on remote cloud hosting providers.
As AI technology becomes more affordable and accessible, startups will need to prioritize real intellectual property over simply embedding AI models into devices. The Orin Nano may not replace Nvidia’s high-end GPUs for training and inferring from large-scale AI models, but it offers a practical solution for running less demanding AI applications.
In a world where AI implementation is increasingly common, startups will need to differentiate themselves with genuine innovation rather than relying on basic AI integration.