Why Video is the First Frontier as AI Learns to Read the Physical World
September 3, 2026 – 9:46 pm
AI Learning and Artificial Intelligence Concept.
Credit: Canva
Summary
Hundreds of millions of cameras are already deployed globally, but most footage still requires a human to identify relevant content. Lumana, founded by ex-Intel computer vision leaders, processes over a billion images daily across 50,000+ cameras using its VIA-1 model. This model learns what is normal for each individual camera and flags deviations, following the principle of "filter before you spend."
Video surveillance may be AI’s natural starting point due to existing infrastructure. While traditional footage relies on human oversight, AI enables meaningful content extraction and search during ongoing events.
Axis Communications Estimates
According to Axis Communications, 562 million surveillance cameras were installed worldwide outside China by the end of 2025, with two-thirds including deep-learning analytics.
Physical AI’s Potential
Companies building physical AI leverage existing camera infrastructure. Lumana, with its founders’ extensive computer vision experience at Intel, is among those betting on this transition. The startup raised $40 million in July 2025 and reported over 50,000 cameras connected to its platform by December, serving Fortune 500 businesses.
Early Impact: Streamlining Security Operations
Lumana’s AI video surveillance systems process more than a billion images daily across 50,000+ cameras. Initially, customers focus on sorting camera issues and optimizing alert distribution. Eventually, operators spend less time monitoring passive video walls and more time investigating flagged events.