AI Craze｜Nvidia Invests in Security Company Verkada to Accelerate Physical AI

Nvidia has made a new investment in Verkada and is stepping up its push for the physical AI market. Through a technology cooperation, the company combines Cosmos’ world foundation model and a physical AI data factory to improve image search accuracy, supplements training data with synthetic data, and advances R&D for physical security and IoT at the same time. Verkada was previously valued at $5.8 billion, with reports saying an IPO could happen within a year. Jensen Huang has predicted that Physical AI will become widely adopted around 2030, with a market size of about $50 trillion.

“Jensen Huang” also known as “Leather Jacket” is steering Nvidia (Nvidia), which has made another investment. In early July, it set its sights on Verkada, a company focused on security cameras and related software—an action seen as a further move to expand the Physical AI market that Jensen Huang favors.

Neither Nvidia nor Verkada has disclosed the investment amount or the number of shares Nvidia obtained. However, many reports that look back at Verkada’s most recent funding round (7 months ago) note that at the time, Verkada’s valuation had already reached $5.8 billion (about HK$45 billion). At that time, it was led by CapitalG, a unit under Alphabet. There are rumors that Verkada will launch an IPO within 1 year.

Verkada brings together Nvidia’s Cosmos world foundation model and its physical AI data factory toolkit. A technical highlight of this solution is video search, which can retrieve specific people, objects, or brief moments from thousands of hours of recorded footage. The data factory generates synthetic videos to fill gaps in training data. Verkada says that with Nvidia’s help, the average accuracy of search has improved by 68%. Verkada will deeply integrate Nvidia’s computing platform to speed up R&D of physical AI technologies such as AI alerts, computer vision, and the Internet of Things (IoT).

As a unicorn in the field of physical smart security, Verkada previously, at the SC West 2026 exhibition, showcased a cloud AI video search solution in a scenario themed “The Louvre Heist,” demonstrating cloud-based AI video search, Person Re-Identification (ReID), license plate recognition, PTZ auto-tracking, integrated access control alerting, and real-time event handling workflows.

Jensen Huang is bullish on Physical AI

Jensen Huang once made a bold prediction that in the future, Physical AI could reshape a market valued at about $50 trillion (about HK$390 trillion), mainly centered on manufacturing and logistics. Physical AI can generally be divided into 3 major areas, including embodied intelligence and robotics, autonomous driving and intelligent transportation, and industrial simulation and digital twin platforms.

Traditional AI mainly processes information such as text and images, while the core of Physical AI lies in enabling intelligent agents to understand space, distance, structure, object attributes, tactile sensation, and changes in force. Jensen Huang said that Physical AI predicts outcomes through basic physics principles, allowing machines to autonomously learn how to interact in the real world. He previously predicted that around 2030, Physical AI—such as robots—will become ubiquitous.

#美終止對伊朗石油制裁豁免