Back to AI TrendsInfrastructure
Niantic’s Spatial Intelligence: Building the Neural Network of the Physical World
Fast Company April 7, 2026
The creator of Pokémon Go is pivoting from gaming to foundational infrastructure by building a 'Large Geospatial Model' to give AI a sense of physical space. For executives, this represents the missing link for 'Embodied AI'—providing the 3D maps robots and autonomous systems need to navigate and understand the real world.
Key Intelligence
- •Apparently, Niantic is using millions of user-submitted scans from its mobile games to train what it calls a 'Large Geospatial Model' (LGM).
- •Did you hear that Niantic is essentially building 'ChatGPT for physical space,' allowing AI to predict the geometry of a room or street it has never seen before.
- •This shift moves the company from a simple game developer to a critical infrastructure provider for the future of robotics and delivery drones.
- •The technology aims to solve the 'localization' problem, helping AR glasses and autonomous devices know exactly where they are within centimeters.
- •Industry insiders view this as a direct challenge to Google and Apple’s dominance in mapping, but with a focus on AI-readiness rather than just human navigation.
- •Niantic's massive dataset is unique because it includes 360-degree views of locations that traditional street-view cars simply cannot reach.
- •The ultimate goal is 'Spatial Intelligence,' a term becoming increasingly popular in Silicon Valley to describe AI that understands and interacts with 3D environments.