Cognex Corporation has signed a definitive agreement to acquire RealSense, a supplier of depth-sensing cameras and vision technology aimed at robotic perception and what the industry now labels Physical AI.

The deal moves Cognex beyond its traditional core of 2D industrial machine vision—factory-floor inspection, identification, and guidance—into 3D depth sensing, the sensing modality that mobile robots, autonomous guided vehicles, and collaborative manipulators depend on for navigation, obstacle avoidance, and grasp planning. RealSense built its position supplying stereo and active-IR depth cameras used in robotics research, logistics automation, and embedded perception systems.

For Cognex customers, the acquisition addresses a practical gap. A 2D smart camera can verify a label or measure a part silhouette, but bin picking, pallet localization, and free-space mapping require calibrated range data. Depth cameras deliver that by projecting infrared patterns and computing per-pixel disparity, producing point clouds accurate to millimeter-level at working distances from tens of centimeters to several meters—performance that turns raw perception into metrologically useful measurement when factory calibration holds.

Neither company disclosed the purchase price, expected closing date, or product roadmap details in the announcement. Cognex characterized RealSense as a market leader in depth-sensing cameras for robotic perception and Physical AI, the term now used for AI systems that act on and within the physical world rather than purely digital data.

The strategic logic is straightforward. Robot deployments in logistics and manufacturing increasingly demand perception stacks that combine 2D identification with 3D spatial awareness, and buyers prefer to source both from one vendor with a common software environment. By folding RealSense's depth camera line into its portfolio, Cognex can offer machine-vision customers a migration path from inspection tasks toward full robotic guidance and spatial measurement.

For integrators, the open questions are the ones that decide projects: whether RealSense cameras will retain their existing SDKs and drivers, whether Cognex's vision tools (VisionPro, In-Sight ecosystems) will gain native point-cloud processing, and what the acquisition means for long-term calibration support and firmware maintenance of the depth units already deployed in the field. Acquisitions of sensor companies have historically led to software-stack consolidation, and integrators will want commitments before specifying the hardware into new cells.

The transaction also signals where Cognex sees industrial measurement heading: toward sensors that do not just verify finished parts but continuously localize and measure in dynamic, unstructured environments. Whether the combined entity can hold RealSense's depth accuracy and reliability through the integration—and keep the developer ecosystem intact—is the question the metrology community will now watch.