Niantic and Spexi are combining drone networks with 3D physical AI to automate high-resolution infrastructure inspections at scale.

Niantic and Spexi are combining drone networks with 3D physical AI to automate high-resolution infrastructure inspections at scale. Rendering an accurate digital twin of physical assets across potentially thousands of acres requires overcoming severe data collection barriers. Standard satellite imagery lacks the resolution required for detailed asset management or energy site analysis. Independent drone operations provide closer views but generate heavy, unstructured visual files that struggle to integrate into wider operational technology platforms. Geometrically-accurate models are vital to manage predictive maintenance pipelines without inflating cloud storage budgets or expanding manual analysis teams. Extracting actionable intelligence from raw visual data currently demands heavy manual processing, pulling engineers away from core diagnostic tasks. Niantic Spatial and Spexi Geospatial have established a partnership to address this gap in the data collection pipeline. The collaboration merges Spexi’s aerial data network with Niantic’s 3D reconstruction technology to produce large-scale 3D intelligence on demand. Under the agreement, users can commission drone captures and retrieve geometrically accurate 3D reconstructions via Niantic’s API. These outputs arrive as 3D Gaussian splats, fully geo-referenced and accessible within the Spexi World platform via a dedicated viewer and measurement tool. Financial impact of metric-scale precision When inspecting extensive utilities or conducting insurance risk assessments, teams require metric-scale precision. Spexi’s network operates with over 10,000 pilots who have mapped more than six million acres at a 2.8 cm resolution. This capture density is ten times sharper than standard satellite feeds. Applying autonomous, standardised flight protocols optimised for machine learning allows energy firms and logistics operators to track physical degradation or site viability remotely. Automated compliance checks and 3D infrastructure measurements reduce site visit frequency, protecting gross margins against rising travel and manual labour expenses. Niantic Spatial has designated Spexi as a preferred drone imagery provider to train its real-world foundation models for physical AI. The system stitches together multiple drone scans to build intelligent models grounded in geometry, highly applicable for spatial simulation and location tracking. Inhi Cho Suh, CEO of Niantic Spatial, said: “For physical AI to work in the real world, it needs a foundation grounded in reality. Combining Spexi’s capture network with our reconstruction technology and real-world models gets us significantly closer to that and delivers real operational value for our customers. “Until now, high-quality 3D reconstruction has largely operated at the scale of an object or building. This partnership takes it to city scale and more.” Architectural integration and edge compute costs Processing 3D Gaussian splats requires careful edge compute provisioning. Plant managers must assess local hardware sustainability before streaming high-volume reconstruction data across industrial networks. Vector database integration becomes necessary to index and query vast spatial dimensions efficiently. When operational technology systems ingest physical AI data, engineers need vector search capabilities to locate specific asset anomalies (e.g. a micro-fracture on a specific cooling tower) within a massive city-scale reconstruction. Connecting these modern spatial systems to legacy architecture adds another layer of friction. Older enterprise asset management databases rely on rigid, relational structures that do not easily accommodate dynamic 3D spatial data. Middleware development is required to translate geo-referenced coordinates from the Spexi World platform into maintenance tickets within legacy enterprise resource planning platforms. Supply chain directors must allocate engineering hours to build these bridges, ensuring that an anomaly detected by physical AI automatically triggers a supply order for replacement parts without manual data entry. Hallucination mitigation in physical AI models Foundation models interpreting physical AI require extensive hallucination mitigation. An AI model identifying rust on a pipeline or assessing structural integrity from drone imagery must not infer damage where shadows, weather anomalies, or lighting artifacts exist. Sticking to geometrically-grounded models reduces false positives, yet validating these outputs against legacy supervisory control and data acquisition systems requires meticulous calibration. Compute costs will also scale proportionally as utility providers transition from periodic manual inspections to continuous, on-demand physical AI processing across thousands of acres. Balancing the processing load between localised edge servers and central cloud repositories will dictate the overall financial viability of these deployments. Bill Lakeland, CEO of Spexi, commented: “Together, Spexi and Niantic Spatial deliver a drone-to-3D pipeline that will redefine the next generation of physical AI, unlocking more accurate, up-to-date, and immersive representations of the built environment. “Partnering with Niantic Spatial means customers can now go from raw imagery to actionable 3D intelligence in one seamless workflow. That’s a step change in what drone data can achieve for real-world applications.” Ensuring that reconstruction pipelines remain calibrated to capture workflows guarantees that every flight yields maximum quality output, preventing data corruption scenarios that waste cloud resources. Delivering an architecture designed for city-scale reconstruction provides industrial leaders with a tangible method for upgrading their asset management systems while tightening operational expenditures. See also: NTT, Kubota, and DOCOMO test links for remote farm robots Want to learn more about the IoT from industry leaders? Check out IoT Tech Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including AI & Big Data Expo and the Cyber Security Expo. Click here for more information.

Total number of new cars licensed in May rose by 32 per cent from 8,403 to 11,063.

The Singapore-based AI-native workforce management and integrated facility management company said the initiative will deploy humanoid robots across commercial environments.

Goal is to accelerate the deployment of humanoid robots and embodied AI in real-world production and service environments.