Niantic Spatial, Inc.
Builds a Large Geospatial Model for 3D reconstruction, localization, and semantic understanding of the physical world.
Updated April 2026
Overview
- Website
- nianticspatial.com
- Founded
- 2025
- Headquarters
- San Francisco, California, USA
- Segment
- Image & Visual Generation
Product overview
Niantic Spatial develops a Large Geospatial Model trained on vast real-world sensor data, providing Reconstruct (high-fidelity 3D models via Gaussian splats), Localize (cm-accurate Visual Positioning for GPS-denied areas), and Understand (spatial reasoning and semantic queries). These enable embodied AI for robots, AR glasses, drones, and enterprises in logistics, entertainment, simulation, and defense. Distinct from competitors with persistent shared spatial context at global scale, fusing 30+ billion posed images for superior real-world AI perception.
Revenue model
Usage-based pricing via SDK: free tiers (limited VPS calls ~10k/month, limited MAU/sessions), then tiered paid plans for higher volumes (per 1,000 VPS calls, MAU, sessions, private locations); enterprise custom licensing, on-demand mapping, and support via contact sales.
Moat
Niantic Spatial's key competitive moat is its proprietary geospatial AR platform, featuring centimeter- to millimeter-level spatial precision for anchoring digital content to the physical world, powered by unique expertise in combining XR, GIS, AI, and real-time localization that competitors struggle to replicate without equivalent data infrastructure and IP. This is reinforced by scale advantages from Niantic's vast proprietary mapping data accumulated through billions of user interactions in games like Pokémon GO, enabling superior pattern recognition, predictive analytics, and applications in warehousing, navigation, and digital twins that create high switching costs for enterprise adopters.
Headwinds
Large Geospatial Models face uncertain market demand and competition from established mapping companies like Google and Apple.