GEODNET builds a positioning layer for physical AI with 22,000 base stations

Robots are getting smarter, but a smart robot that doesn’t know exactly where it is remains a liability. GEODNET is betting that the fix is a global grid of cheap base stations, and the network now counts more than 22,000 of them. The decentralized project says it delivers centimeter-level positioning data to AI-driven devices. It serves over 150 enterprise customers and booked revenue of $2.8 million in Q3, a rare instance of a crypto infrastructure network pointing to paying clients rather than just a token chart. What GEODNET actually does GEODNET’s answer is real-time kinematic corrections, usually shortened to RTK. Ground-based reference stations sit at known, fixed locations and monitor satellite signals. They then broadcast correction data so nearby devices can sharpen their own position estimates. According to the project, the result is ±1 cm accuracy. The network spans 174 countries and had passed 22,000 base stations as of September 30, 2026. It reportedly runs hundreds of thousands of GB in operational RTK data streams. Customers span drones, robotics, autonomous vehicles, precision agriculture, and surveying. The DePIN model and the token mechanics GEODNET runs on what ...