Nvidia has announced the Jetson Orin Nano 2, a small computer designed to run artificial intelligence directly on robots, drones, cameras and other machines rather than in a distant data center.
According to a Techmeme summary of reporting by Eugene Demaitre at The Robot Report, Nvidia says the device doubles inference performance — the speed at which an AI model turns incoming data into decisions — and delivers 78 TOPS of AI compute alongside an eight-core Arm processor. TOPS, or trillions of operations per second, is the industry's rough shorthand for raw AI horsepower.
Unite.ai reports that the announcement came on August 25, 2026, and describes the Orin Nano 2 as the successor to Nvidia's entry-level robotics computer, doubling inference performance over the previous Jetson Orin Nano Super while holding the same price. Stock Titan notes that automotive supplier Aptiv is among the companies supporting the new module.
The framing from The Robot Report gets at the reason this matters: as AI models become more efficient, more devices can become autonomous — but developers need compact, energy-conscious hardware to run them. A machine that thinks locally doesn't have to wait on a network round trip to a cloud server, which matters when the decision is whether to stop before hitting something.
The entry-level tier is also where the hobbyists, university labs and startups live. Keeping the price flat while doubling throughput lowers the cost of experimenting with more capable models, which is how a lot of robotics work begins.
One caveat worth stating plainly: the performance figures here come from Nvidia's own announcement, relayed through trade coverage, not from independent benchmarks.
It matters because cheaper on-device AI compute is the quiet precondition for the wave of autonomous machines companies keep promising.