SDG&E and Qualcomm Deploy Edge AI for Real-Time Wildfire Detection

San Diego Gas & Electric, Qualcomm Technologies, and the University of California San Diego’s Scripps Institution of Oceanography have officially launched Edge Alert Sentinel, a specialized artificial intelligence system for wildfire monitoring. The deployment, situated on Mt. Palomar in Southern California, processes wind, weather, and environmental telemetry directly at the point of collection.

The architecture relies on a ruggedized edge AI gateway powered by the Qualcomm Dragonwing IQ9 processor. This hardware integrates a dedicated neural-processing unit capable of executing up to 100 trillion operations per second. By shifting computation from remote data centers to the edge, the system reduces latency during critical weather events.

Operational data flows through an MLOps framework provided by Edge Impulse to forecast conditions that could threaten grid infrastructure. SDG&E, a subsidiary of Sempra, integrates these insights into its existing control center via a private cellular network. The collaboration combines Sempra’s grid infrastructure expertise with Qualcomm’s edge-computing hardware and Scripps’s observational data.

The MLOps pipeline utilizes Edge Impulse to manage the lifecycle of models deployed on the Dragonwing IQ9 hardware. Engineers train these models on historical weather patterns to identify specific environmental signatures that precede wildfire ignition. Once deployed, the models perform real-time inference on sensor data, filtering out noise to isolate high-risk anomalies.

This localized processing capability ensures that the system remains functional even when wide-area network connectivity is compromised by extreme weather. By minimizing the need to transmit raw data to centralized servers, the architecture preserves bandwidth for critical alert signals. The system effectively turns remote sensor nodes into intelligent, autonomous monitoring stations.

Scott Crider, President of SDG&E, stated,

For nearly two decades, our region has avoided a catastrophic electrically caused wildfire because we chose to lead early and never stop looking ahead.

The current phase focuses on evaluating system performance during the upcoming Public Safety Power Shutoff season. Future iterations will include a broader rollout across additional sites starting in 2027.

The technical roadmap also includes the integration of automated infrastructure inspections using autonomous aerial operations. These efforts aim to minimize response times during rapidly changing environmental conditions where data transmission delays previously hindered emergency management. The project represents a shift toward localized AI inference for high-stakes infrastructure monitoring.

The reliance on edge-based inference addresses a fundamental bottleneck in traditional environmental monitoring systems. By performing data analysis locally, the system ensures that the neural-processing unit can trigger alerts even when wide-area network connectivity is degraded by extreme weather. This architecture provides a more resilient alternative to cloud-dependent monitoring solutions.

The integration of MLOps at the edge requires precise model optimization to ensure the Dragonwing IQ9 processor maintains high throughput without exceeding power constraints. This deployment serves as a practical case study for the application of high-performance neural processing in remote, rugged environments. The success of the Palomar Mountain pilot will dictate the scalability of the model for larger grid-monitoring applications.

Sempra reported first-quarter 2026 adjusted earnings per share of $1.51, exceeding analyst expectations of $1.49, according to the company’s recent financial disclosures. The firm posted $3.66 billion in revenue, falling short of the $4.1 billion anticipated by market observers. Sempra Infrastructure is currently undergoing a leadership transition as Bob Patel prepares to assume the CEO role following a KKR-led consortium’s acquisition of a 65% stake in the subsidiary.

Stakeholders will monitor the system’s accuracy in identifying wildfire precursors during the upcoming fire season. The transition from pilot testing to a wider deployment in 2027 will serve as the primary indicator of the technology’s effectiveness in real-world, high-stakes scenarios. Analysts will evaluate whether the edge-computing approach provides a measurable reduction in grid-related fire risk compared to legacy monitoring methods.