Verkada says it has entered a technical collaboration with NVIDIA aimed at developing and deploying what the companies describe as “physical AI” for the built environment, with NVIDIA also taking an investment stake in the physical security technology provider.
The announcement follows a strategic investment from Alphabet’s CapitalG late last year, according to Verkada. The company did not disclose the size of NVIDIA’s investment.
Verkada said the collaboration focuses on improving AI-driven video analytics, including video search, multimodal embeddings and vector retrieval for semantic search, and the use of synthetic data generation to augment training datasets.
The company said it is using NVIDIA Cosmos world foundation models and the NVIDIA Physical AI Data Factory to accelerate model training and inference across more than 2.4 million devices deployed in 170 countries. Verkada also claimed the collaboration has produced a 68% improvement in mean average precision (mAP) for its AI-powered search, which it said reduces investigation time for customers.
“Verkada has been building and deploying Physical AI before the term existed. With our footprint of more than 2.4 million devices across 170 countries and 30,000 organizations, we’ve proven that the built environment is one of the largest beneficiaries of AI,” said Filip Kaliszan, co-founder and CEO of Verkada. “Working with NVIDIA supercharges what we’ve spent nearly a decade building: AI that keeps students safe in schools, protects workers on factory floors, helps retailers prevent theft, and enables organizations to operate more efficiently.”
Verkada said it is also developing a multi-model search agent architecture and exploring reasoning models intended to address “complex, unstructured real-world scenarios,” including health and safety incidents on manufacturing floors and retail shrinkage detection.
The deal highlights the growing role of large-scale compute and synthetic data techniques in physical security analytics, as vendors seek faster model development and improved accuracy across large device fleets. It also underscores the increased interest from major AI infrastructure providers in the physical security market as video and sensor data volumes expand.
